framework,version,device,op_name,kernel_source,moe_dtype,num_tokens,hidden_size,inter_size,topk,num_experts,moe_tp_size,moe_ep_size,distribution,latency
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,1,balanced,0.1364959975083669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,1,balanced,0.20905599991480509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,1,balanced,0.39665599664052326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,1,balanced,0.19955732425053915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,1,balanced,0.12441600362459819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,1,balanced,0.5791946649551392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,1,balanced,0.34253867467244464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,1,balanced,0.20011200507481894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,1,balanced,0.5928053458531698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,1,balanced,0.03400533397992452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,1,balanced,0.3149440089861552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,1,balanced,0.3328373432159424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,1,balanced,0.5913120110829672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,1,balanced,0.30644800265630084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,1,balanced,0.0347626656293869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,1,balanced,0.3352213303248088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,1,balanced,0.6070026556650797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,1,balanced,0.34112000465393066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,1,balanced,0.03642133375008901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,1,balanced,0.3492106596628825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,1,balanced,0.6439786752065023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,1,balanced,0.362885316212972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,1,balanced,0.32052799065907794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,1,balanced,0.04021333406368891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,1,balanced,0.6112053394317627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,1,balanced,0.33590932687123615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,1,balanced,0.35871466000874835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,1,balanced,0.049327999353408813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,1,balanced,0.5997386773427328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,1,balanced,0.3487893342971802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,1,balanced,0.3307093381881714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,1,balanced,0.07182399928569794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,1,balanced,0.6158186594645182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,1,balanced,0.3322239915529887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,1,balanced,0.34727998574574787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,1,balanced,0.0655680000782013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,1,balanced,0.6465813318888346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,1,balanced,0.339903990427653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,1,balanced,0.36241066455841064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,1,balanced,0.07330666482448578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,1,balanced,0.59826131661733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,1,balanced,0.311408003171285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,1,balanced,0.31174399455388385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,1,balanced,0.06270933151245117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,1,balanced,0.581712007522583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,1,balanced,0.306549330552419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,1,balanced,0.31842132409413654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,1,balanced,0.06798933446407318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,1,balanced,0.5331413348515829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,1,balanced,0.30832000573476154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,1,balanced,0.3052906592686971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,1,balanced,0.07737066845099132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,1,balanced,0.5130826632181803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,1,balanced,0.302021324634552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,1,balanced,0.3018079996109009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,1,balanced,0.07341866691907246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,1,balanced,0.5021493434906006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,1,balanced,0.2686240077018738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,1,balanced,0.3018186688423157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,1,balanced,0.078575998544693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,1,balanced,0.5094613234202067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,1,balanced,0.2573546568552653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,1,balanced,0.3054826656977336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,1,balanced,0.08795733253161113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,1,balanced,0.9196906884511312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,1,balanced,0.23702933390935263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,1,balanced,0.5087999900182089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,1,balanced,0.17198934157689413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,1,balanced,0.8846720059712728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,1,balanced,0.489898681640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,1,balanced,0.18893865744272867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,1,balanced,0.47837332884470624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,1,balanced,0.23906133572260538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,1,balanced,0.36629335085550946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,1,balanced,1.2651306788126628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,1,balanced,0.6551786661148071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,1,balanced,0.13338133692741394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,1,balanced,0.5071039994557699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,1,balanced,1.5988693237304688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,1,balanced,0.8152799606323242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,1,balanced,0.15744533141454062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,1,balanced,0.651472012201945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,1,balanced,1.1483360131581624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,1,balanced,2.305311997731527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,1,balanced,0.22021333376566568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,1,balanced,0.9319732983907064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,1,balanced,0.2803679903348287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,1,balanced,1.4758186340332031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,1,balanced,2.9371681213378906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,1,balanced,1.2081387042999268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,1,balanced,0.3941226800282796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,1,balanced,2.113482634226481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,1,balanced,4.160229365030925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,1,balanced,1.7808747291564941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,1,balanced,0.5033226807912191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,1,balanced,2.7391252517700195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,1,balanced,5.230736096700032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,1,balanced,0.7392106850941976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,1,balanced,2.349829355875651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,1,balanced,0.9626879692077637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,1,balanced,3.978954633076986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,1,balanced,7.468522389729817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,1,balanced,3.4640480677286782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,1,balanced,1.4222240447998047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,1,balanced,5.23525873819987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,1,balanced,4.609706560770671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,1,balanced,9.767754872639975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,1,balanced,1.8552533785502117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,1,balanced,6.5086930592854815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,1,balanced,12.025717417399088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,1,balanced,5.749077479044597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,1,balanced,2.3152106602986655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,1,balanced,10.326218922932943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,1,balanced,3.6693814595540366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,1,balanced,18.910602569580078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,1,balanced,9.208607991536459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,1,balanced,20.474815368652344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,1,balanced,7.284789403279622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,1,balanced,36.50795237223307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,1,balanced,18.2971674601237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,1,power_law_1.01,0.13696639537811278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,1,power_law_1.01,0.22576639652252198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,1,power_law_1.01,0.1668352007865906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,1,power_law_1.01,0.2735487937927246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,1,power_law_1.01,0.19962879419326782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,1,power_law_1.01,0.37233920097351075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,1,power_law_1.01,0.3273920059204102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,1,power_law_1.01,0.593932819366455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,1,power_law_1.01,0.3288383960723877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,1,power_law_1.01,0.611679983139038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,1,power_law_1.01,0.12005120515823364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,1,power_law_1.01,0.3375808000564575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,1,power_law_1.01,0.5750080108642578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,1,power_law_1.01,0.17310719490051268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,1,power_law_1.01,0.340447998046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,1,power_law_1.01,0.5950528144836426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,1,power_law_1.01,0.18815360069274903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,1,power_law_1.01,0.3648384094238281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,1,power_law_1.01,0.598470401763916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,1,power_law_1.01,0.031699201464653014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,1,power_law_1.01,0.3604032039642334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,1,power_law_1.01,0.3411967992782593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,1,power_law_1.01,0.6014143943786621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,1,power_law_1.01,0.03223040103912354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,1,power_law_1.01,0.3353600025177002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,1,power_law_1.01,0.5989888191223145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,1,power_law_1.01,0.34225919246673586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,1,power_law_1.01,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,1,power_law_1.01,0.3335616111755371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,1,power_law_1.01,0.5983232021331787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,1,power_law_1.01,0.328767991065979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,1,power_law_1.01,0.03868159949779511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,1,power_law_1.01,0.3311295986175537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,1,power_law_1.01,0.6165120124816894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,1,power_law_1.01,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,1,power_law_1.01,0.33431038856506345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,1,power_law_1.01,0.37223680019378663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,1,power_law_1.01,0.6419199943542481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,1,power_law_1.01,0.06998400092124939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,1,power_law_1.01,0.3428992033004761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,1,power_law_1.01,0.38777599334716795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,1,power_law_1.01,0.6675136089324951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,1,power_law_1.01,0.06170880198478699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,1,power_law_1.01,0.33420801162719727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,1,power_law_1.01,0.39831678867340087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,1,power_law_1.01,0.7404863834381104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,1,power_law_1.01,0.07043200135231018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,1,power_law_1.01,0.3376192092895508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,1,power_law_1.01,0.4107071876525879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,1,power_law_1.01,0.7461184024810791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,1,power_law_1.01,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,1,power_law_1.01,0.3264319896697998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,1,power_law_1.01,0.06565120220184326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,1,power_law_1.01,0.4586304187774658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,1,power_law_1.01,0.7920383930206298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,1,power_law_1.01,0.32103040218353274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,1,power_law_1.01,0.07363200187683105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,1,power_law_1.01,0.5481279850006103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,1,power_law_1.01,0.9902463912963867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,1,power_law_1.01,0.33930881023406984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,1,power_law_1.01,0.07178879976272583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,1,power_law_1.01,0.6615424156188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,1,power_law_1.01,1.1801664352416992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,1,power_law_1.01,0.34245760440826417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,1,power_law_1.01,0.076665598154068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,1,power_law_1.01,0.8006976127624512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,1,power_law_1.01,1.530508804321289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,1,power_law_1.01,0.38278400897979736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,1,power_law_1.01,0.08739839792251587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,1,power_law_1.01,0.9681856155395507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,1,power_law_1.01,1.8639808654785157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,1,power_law_1.01,0.38530559539794923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,1,power_law_1.01,0.1704192042350769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,1,power_law_1.01,1.3161855697631837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,1,power_law_1.01,2.5320959091186523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,1,power_law_1.01,0.41032958030700684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,1,power_law_1.01,0.1883072018623352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,1,power_law_1.01,1.6315263748168944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,1,power_law_1.01,3.1770751953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,1,power_law_1.01,0.49619197845458984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,1,power_law_1.01,0.2384511947631836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,1,power_law_1.01,2.278495979309082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,1,power_law_1.01,4.304588699340821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,1,power_law_1.01,0.13227519989013672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.01,0.5345344066619873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,1,power_law_1.01,2.880518341064453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.01,0.1549631953239441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,1,power_law_1.01,5.485894393920899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.01,0.738976001739502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,1,power_law_1.01,4.151065444946289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.01,0.21933441162109374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.01,0.8518272399902344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,1,power_law_1.01,7.753740692138672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.01,0.2784832000732422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,1,power_law_1.01,5.396486282348633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.01,1.1194560050964355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.01,0.3954432010650635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,1,power_law_1.01,9.926976013183594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.01,1.3852352142333983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,1,power_law_1.01,6.678662109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.01,0.5037951946258545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.01,1.940332794189453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,1,power_law_1.01,12.070329284667968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.01,0.7355711936950684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,1,power_law_1.01,10.429068756103515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.01,2.488422393798828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.01,0.9631999969482422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.01,3.638719940185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.01,1.4131967544555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,1,power_law_1.01,18.783180236816406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,1,power_law_1.01,20.681106567382812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.01,1.8601408004760742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.01,4.833881759643555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.01,2.3213375091552733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.01,5.884096145629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,1,power_law_1.01,36.771417236328126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,1,power_law_1.01,3.674918365478516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,1,power_law_1.01,9.259430694580079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,1,power_law_1.01,7.297132873535157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,1,power_law_1.01,18.425389099121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,1,power_law_1.2,0.2287424087524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,1,power_law_1.2,0.2758336067199707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,1,power_law_1.2,0.12499200105667115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,1,power_law_1.2,0.35601279735565183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,1,power_law_1.2,0.1567296028137207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,1,power_law_1.2,0.5726336002349853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,1,power_law_1.2,0.11886719465255738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,1,power_law_1.2,0.21628799438476562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,1,power_law_1.2,0.5936639785766602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,1,power_law_1.2,0.14071040153503417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,1,power_law_1.2,0.35532801151275634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,1,power_law_1.2,0.6227776050567627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,1,power_law_1.2,0.18522239923477174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,1,power_law_1.2,0.6015103816986084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,1,power_law_1.2,0.3502847909927368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,1,power_law_1.2,0.34591360092163087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,1,power_law_1.2,0.5944831848144532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,1,power_law_1.2,0.3352960109710693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,1,power_law_1.2,0.3376255989074707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,1,power_law_1.2,0.031385600566864014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,1,power_law_1.2,0.5871935844421386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,1,power_law_1.2,0.34926719665527345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,1,power_law_1.2,0.3373183965682983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,1,power_law_1.2,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,1,power_law_1.2,0.617523193359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,1,power_law_1.2,0.34953598976135253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,1,power_law_1.2,0.33844480514526365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,1,power_law_1.2,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,1,power_law_1.2,0.6274559974670411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,1,power_law_1.2,0.3361855983734131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,1,power_law_1.2,0.3261631965637207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,1,power_law_1.2,0.03761920034885406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,1,power_law_1.2,0.6456575870513916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,1,power_law_1.2,0.3307775974273682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,1,power_law_1.2,0.32714879512786865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,1,power_law_1.2,0.04731520116329193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,1,power_law_1.2,0.6056640148162842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,1,power_law_1.2,0.3415488004684448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,1,power_law_1.2,0.3404288053512573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,1,power_law_1.2,0.06972799897193908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,1,power_law_1.2,0.6873856067657471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,1,power_law_1.2,0.3307584047317505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,1,power_law_1.2,0.3321727991104126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,1,power_law_1.2,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,1,power_law_1.2,0.7368576049804687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,1,power_law_1.2,0.3426431894302368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,1,power_law_1.2,0.35507841110229493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,1,power_law_1.2,0.07120640277862549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,1,power_law_1.2,0.7625664234161377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,1,power_law_1.2,0.35208959579467775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,1,power_law_1.2,0.3660671949386597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,1,power_law_1.2,0.060838401317596436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,1,power_law_1.2,0.78438401222229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,1,power_law_1.2,0.35961599349975587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,1,power_law_1.2,0.42060160636901855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,1,power_law_1.2,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,1,power_law_1.2,0.9689408302307129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,1,power_law_1.2,0.38078720569610597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,1,power_law_1.2,0.40265598297119143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,1,power_law_1.2,0.07304319739341736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,1,power_law_1.2,1.193612766265869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,1,power_law_1.2,0.3852992057800293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,1,power_law_1.2,0.44797439575195314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,1,power_law_1.2,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,1,power_law_1.2,1.505286407470703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,1,power_law_1.2,0.42997121810913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,1,power_law_1.2,0.5414591789245605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,1,power_law_1.2,0.07672960162162781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,1,power_law_1.2,1.8367424011230469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,1,power_law_1.2,0.466431999206543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,1,power_law_1.2,0.6400000095367432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,1,power_law_1.2,0.08686720132827759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.2,0.5699584007263183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,1,power_law_1.2,2.5149503707885743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,1,power_law_1.2,0.16919039487838744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,1,power_law_1.2,0.8066176414489746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,1,power_law_1.2,0.188646399974823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.2,0.7044928073883057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,1,power_law_1.2,0.9546175956726074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,1,power_law_1.2,3.155955123901367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,1,power_law_1.2,0.2371903896331787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.2,0.815071964263916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,1,power_law_1.2,1.2705280303955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,1,power_law_1.2,4.332089614868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,1,power_law_1.2,0.13188480138778685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.2,1.0946687698364257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,1,power_law_1.2,1.6274112701416015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,1,power_law_1.2,5.465177536010742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.2,0.1551424026489258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.2,1.3685952186584474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,1,power_law_1.2,2.27127685546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.2,0.21982080936431886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,1,power_law_1.2,7.744998168945313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.2,1.9532800674438477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,1,power_law_1.2,2.908121681213379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.2,0.2792831897735596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.2,2.509343910217285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,1,power_law_1.2,10.012947082519531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.2,0.3994879961013794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,1,power_law_1.2,4.152691268920899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.2,3.613209533691406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.2,0.5000256061553955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,1,power_law_1.2,5.407251358032227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,1,power_law_1.2,12.09180145263672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.2,0.7401792049407959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.2,4.709708786010742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,1,power_law_1.2,6.643180847167969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.2,0.9628479957580567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.2,5.864038467407227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,1,power_law_1.2,18.822610473632814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.2,1.4175999641418457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,1,power_law_1.2,10.51288299560547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.2,1.85533447265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,1,power_law_1.2,9.193158721923828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.2,2.3163263320922853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,1,power_law_1.2,36.95350952148438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,1,power_law_1.2,20.638201904296874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,1,power_law_1.2,18.235302734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,1,power_law_1.2,3.678444671630859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,1,power_law_1.2,7.304691314697266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,1,balanced,0.34656532605489093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,1,balanced,0.20466132958730063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,1,balanced,0.5339200099309286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,1,balanced,0.3445546627044678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,1,balanced,0.9266773064931234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,1,balanced,0.5466773509979248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,1,balanced,0.9717973073323568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,1,balanced,0.551471988360087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,1,balanced,1.100053310394287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,1,balanced,0.5467199881871542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,1,balanced,1.016976038614909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,1,balanced,0.5558773279190063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,1,balanced,0.04855999847253164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,1,balanced,1.0578933556874592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,1,balanced,0.1888213356335958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,1,balanced,0.5642933448155721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,1,balanced,0.04905066887537638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,1,balanced,0.30244799455006915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,1,balanced,1.0044533411661785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,1,balanced,0.5437599817911783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,1,balanced,0.04931733508904775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,1,balanced,0.5389013290405273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,1,balanced,0.9531253178914388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,1,balanced,0.5965760151545206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,1,balanced,0.05922133227189382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,1,balanced,0.521071990331014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,1,balanced,1.0054933230082195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,1,balanced,0.5681920051574707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,1,balanced,0.0928053359190623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,1,balanced,0.52074134349823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,1,balanced,0.9624959627787272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,1,balanced,0.5283679962158203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,1,balanced,0.14917332927385965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,1,balanced,0.5205173492431641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,1,balanced,0.9225707054138184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,1,balanced,0.12308800220489502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,1,balanced,0.502837340037028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,1,balanced,0.5393600066502889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,1,balanced,0.9392639795939127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,1,balanced,0.15015467007954916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,1,balanced,0.5029919942220052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,1,balanced,0.547327995300293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,1,balanced,0.8709812959035238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,1,balanced,0.10955733060836792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,1,balanced,0.5208319822947184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,1,balanced,0.521941343943278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,1,balanced,0.8510133425394694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,1,balanced,0.12328533331553142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,1,balanced,0.5318826834360758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,1,balanced,0.46718398729960126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,1,balanced,0.15731733043988547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,1,balanced,0.8308320045471191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,1,balanced,0.5421813329060873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,1,balanced,0.4628640015920003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,1,balanced,0.11821867028872173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,1,balanced,0.8314186731974283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,1,balanced,0.4924693504969279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,1,balanced,0.4721813201904297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,1,balanced,0.13426666458447775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,1,balanced,0.5131253401438395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,1,balanced,1.5584534009297688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,1,balanced,0.8239626884460449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,1,balanced,0.1728960076967875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,1,balanced,0.4649973313013713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,1,balanced,1.4865759213765461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,1,balanced,0.2634879946708679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,1,balanced,0.8104639848073324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,1,balanced,0.4250346819559733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,1,balanced,0.3073226610819499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,1,balanced,2.1150879859924316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,1,balanced,0.39317333698272705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,1,balanced,1.1278080145517986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,1,balanced,0.3963306744893392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,1,balanced,0.3457706769307454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,1,balanced,2.788271903991699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,1,balanced,1.4515573183695476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,1,balanced,0.20893865823745728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,1,balanced,0.7692106564839681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,1,balanced,0.26808534065882367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,1,balanced,4.119872093200684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,1,balanced,0.5754666725794474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,1,balanced,2.0521814028422036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,1,balanced,0.3746773401896159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,1,balanced,0.8135573069254557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,1,balanced,5.3862559000651045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,1,balanced,2.5789599418640137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,1,balanced,0.49556267261505127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,1,balanced,1.04202667872111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,1,balanced,0.7071839968363444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,1,balanced,7.8410078684488935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,1,balanced,1.491909344991048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,1,balanced,3.6235198974609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,1,balanced,0.933194637298584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,1,balanced,1.9393386840820312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,1,balanced,10.179248174031576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,1,balanced,4.5944318771362305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,1,balanced,1.3689653078715007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,1,balanced,2.8434133529663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,1,balanced,1.808047930399577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,1,balanced,6.5730133056640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,1,balanced,3.741349220275879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,1,balanced,14.284506479899088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,1,balanced,2.6857385635375977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,1,balanced,5.517669041951497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,1,balanced,8.590773264567057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,1,balanced,17.965423583984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,1,balanced,3.5614614486694336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,1,balanced,7.31437873840332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,1,balanced,21.741200764973957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,1,balanced,10.544314702351889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,1,balanced,4.438378651936849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,1,balanced,9.120698928833008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,1,balanced,16.57976531982422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,1,balanced,33.20617167154948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,1,balanced,7.073370615641276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,1,balanced,14.62350336710612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,1,balanced,14.086991628011068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,1,balanced,32.95135498046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,1,balanced,64.61173502604167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,1,balanced,29.30811309814453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,1,power_law_1.01,0.3337599992752075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,1,power_law_1.01,0.40972161293029785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,1,power_law_1.01,0.5170432090759277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,1,power_law_1.01,0.9320192337036133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,1,power_law_1.01,0.21169919967651368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,1,power_law_1.01,0.9149503707885742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,1,power_law_1.01,0.21006720066070556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,1,power_law_1.01,0.23243520259857178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,1,power_law_1.01,1.0131967544555665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,1,power_law_1.01,0.2607872009277344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,1,power_law_1.01,0.04689280092716217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,1,power_law_1.01,0.29538559913635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,1,power_law_1.01,0.9674304008483887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,1,power_law_1.01,0.30954880714416505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,1,power_law_1.01,0.048665601015090945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,1,power_law_1.01,0.9951359748840332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,1,power_law_1.01,0.5707456111907959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,1,power_law_1.01,0.5924160003662109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,1,power_law_1.01,0.04693120121955872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,1,power_law_1.01,0.9791487693786621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,1,power_law_1.01,0.5311744213104248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,1,power_law_1.01,0.5281280040740967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,1,power_law_1.01,0.05761280059814453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,1,power_law_1.01,0.9850496292114258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,1,power_law_1.01,0.5264768123626709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,1,power_law_1.01,0.5542335987091065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,1,power_law_1.01,0.08947200179100037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,1,power_law_1.01,0.9678336143493652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,1,power_law_1.01,0.5456064224243165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,1,power_law_1.01,0.5886591911315918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,1,power_law_1.01,0.14506239891052247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,1,power_law_1.01,1.058579158782959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,1,power_law_1.01,0.5609600067138671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,1,power_law_1.01,0.11442559957504272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,1,power_law_1.01,1.03372163772583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,1,power_law_1.01,0.5640448093414306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,1,power_law_1.01,0.14563839435577391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,1,power_law_1.01,1.1041215896606444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,1,power_law_1.01,0.5764416217803955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,1,power_law_1.01,0.10524799823760986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,1,power_law_1.01,1.1615936279296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,1,power_law_1.01,0.5521088123321534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,1,power_law_1.01,0.11717760562896729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,1,power_law_1.01,1.1476032257080078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,1,power_law_1.01,0.598476791381836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,1,power_law_1.01,1.3539648056030273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,1,power_law_1.01,0.15047680139541625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,1,power_law_1.01,0.594323205947876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,1,power_law_1.01,0.11528960466384888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,1,power_law_1.01,0.5937600135803223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,1,power_law_1.01,1.7009920120239257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,1,power_law_1.01,0.6569856166839599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,1,power_law_1.01,0.13012479543685912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,1,power_law_1.01,1.9819456100463868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,1,power_law_1.01,0.693235206604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,1,power_law_1.01,0.16750719547271728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,1,power_law_1.01,2.6311744689941405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,1,power_law_1.01,0.7293824195861817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,1,power_law_1.01,0.263481593132019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,1,power_law_1.01,3.2179969787597655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,1,power_law_1.01,0.9559616088867188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,1,power_law_1.01,0.30690560340881345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,1,power_law_1.01,4.486809539794922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,1,power_law_1.01,1.0369855880737304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,1,power_law_1.01,0.39314560890197753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,1,power_law_1.01,5.622419357299805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,1,power_law_1.01,1.357868766784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,1,power_law_1.01,0.2069119930267334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,1,power_law_1.01,7.708211517333984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,1,power_law_1.01,1.6566335678100585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.01,0.26725120544433595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.01,0.3736511945724487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,1,power_law_1.01,2.2744640350341796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,1,power_law_1.01,9.938124847412109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.01,0.4942016124725342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,1,power_law_1.01,2.7986560821533204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,1,power_law_1.01,14.025010681152343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.01,0.7074304103851319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,1,power_law_1.01,3.8186111450195312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.01,0.9322879791259766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,1,power_law_1.01,4.9140159606933596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,1,power_law_1.01,18.031173706054688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.01,1.371180820465088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,1,power_law_1.01,6.840383911132813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.01,1.802886390686035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,1,power_law_1.01,21.992083740234374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,1,power_law_1.01,8.954496002197265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.01,2.6748992919921877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.01,3.547129440307617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,1,power_law_1.01,11.145875549316406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.01,4.42248306274414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,1,power_law_1.01,33.433370971679686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,1,power_law_1.01,16.99541778564453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,1,power_law_1.01,7.043711853027344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,1,power_law_1.01,14.036773681640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,1,power_law_1.01,33.25779724121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,1,power_law_1.01,64.40037231445312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,1,power_law_1.2,0.33325440883636476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,1,power_law_1.2,0.43904638290405273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,1,power_law_1.2,0.5154111862182618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,1,power_law_1.2,0.9772607803344726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,1,power_law_1.2,1.003059196472168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,1,power_law_1.2,0.17938560247421265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,1,power_law_1.2,0.20904319286346434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,1,power_law_1.2,1.0135295867919922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,1,power_law_1.2,0.2381056070327759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,1,power_law_1.2,0.2619712114334106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,1,power_law_1.2,1.0213567733764648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,1,power_law_1.2,0.3418303966522217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,1,power_law_1.2,0.2521984100341797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,1,power_law_1.2,0.9824000358581543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,1,power_law_1.2,0.5350527763366699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,1,power_law_1.2,0.5138688087463379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,1,power_law_1.2,1.0139391899108887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,1,power_law_1.2,0.6062335968017578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,1,power_law_1.2,0.5404352188110352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,1,power_law_1.2,0.9698944091796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,1,power_law_1.2,0.581657600402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,1,power_law_1.2,0.5353216171264649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,1,power_law_1.2,0.9941632270812988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,1,power_law_1.2,0.5739007949829101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,1,power_law_1.2,0.5437183856964112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,1,power_law_1.2,1.0599103927612306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,1,power_law_1.2,0.5755712032318115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,1,power_law_1.2,0.5399744033813476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,1,power_law_1.2,1.003865623474121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,1,power_law_1.2,0.5585023880004882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,1,power_law_1.2,0.5332543849945068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,1,power_law_1.2,1.0637056350708007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,1,power_law_1.2,0.5425087928771972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,1,power_law_1.2,0.5203328132629395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,1,power_law_1.2,1.1931776046752929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,1,power_law_1.2,0.5462912082672119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,1,power_law_1.2,0.526636791229248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,1,power_law_1.2,1.2310400009155273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,1,power_law_1.2,0.5635200023651123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,1,power_law_1.2,0.5037951946258545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,1,power_law_1.2,0.5768832206726074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,1,power_law_1.2,1.4236672401428223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,1,power_law_1.2,0.5948991775512695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,1,power_law_1.2,0.6075071811676025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,1,power_law_1.2,1.745542335510254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,1,power_law_1.2,0.5710976123809814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,1,power_law_1.2,0.6587456226348877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,1,power_law_1.2,2.005299186706543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,1,power_law_1.2,0.5502912044525147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,1,power_law_1.2,0.6897535800933838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,1,power_law_1.2,2.622630310058594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,1,power_law_1.2,0.6088831901550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,1,power_law_1.2,0.7459199905395508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,1,power_law_1.01,0.5268288135528565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,1,power_law_1.2,0.6529407978057862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,1,power_law_1.2,3.273516845703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,1,power_law_1.2,0.9498111724853515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,1,power_law_1.01,0.5429952144622803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,1,power_law_1.2,0.7727551937103272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,1,power_law_1.2,4.357747268676758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,1,power_law_1.2,1.0915776252746583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,1,power_law_1.01,0.5300288200378418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.2,0.8951871871948243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,1,power_law_1.2,5.596435165405273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,1,power_law_1.01,0.5324480056762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,1,power_law_1.2,1.3704319953918458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.2,1.0845888137817383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,1,power_law_1.2,7.886790466308594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,1,power_law_1.01,0.5180672168731689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,1,power_law_1.2,1.6884479522705078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.2,1.3355327606201173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,1,power_law_1.01,0.5512447834014893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,1,power_law_1.2,2.1987583160400392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,1,power_law_1.2,10.44039077758789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.2,1.7625343322753906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,1,power_law_1.01,0.5782464027404786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,1,power_law_1.2,2.7888383865356445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.2,2.2473215103149413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,1,power_law_1.2,14.570098876953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,1,power_law_1.01,0.5965248107910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,1,power_law_1.2,3.8705726623535157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.2,3.08734073638916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,1,power_law_1.01,0.6148223876953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,1,power_law_1.2,4.904998397827148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.2,4.004076766967773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,1,power_law_1.2,18.10401306152344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,1,power_law_1.01,0.6833600044250489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,1,power_law_1.01,0.8040063858032227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,1,power_law_1.2,6.82938232421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.2,5.8230335235595705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.01,0.8856255531311035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,1,power_law_1.2,22.07788848876953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.2,7.583360290527343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,1,power_law_1.2,8.855110168457031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.01,1.1212096214294434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.01,1.2920191764831543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.2,9.404633331298829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,1,power_law_1.2,10.981062316894532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,1,power_law_1.2,33.19178771972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.01,1.807379150390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,1,power_law_1.2,14.751455688476563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.01,2.293164825439453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,1,power_law_1.2,16.96611785888672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.01,3.15914249420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.01,4.081017684936524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,1,power_law_1.2,29.441510009765626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,1,power_law_1.2,33.03337707519531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,1,power_law_1.2,64.66967163085937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.01,5.855104064941406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.01,7.686361694335938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.01,9.423673248291015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,1,power_law_1.01,14.872627258300781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,1,power_law_1.01,29.358258056640626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,1,power_law_1.2,0.045484799146652224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,1,power_law_1.2,0.04754559993743897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,1,power_law_1.2,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,1,power_law_1.2,0.05578240156173706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,1,power_law_1.2,0.08727040290832519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,1,power_law_1.2,0.14491519927978516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,1,power_law_1.2,0.11408640146255493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,1,power_law_1.2,0.14536319971084594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,1,power_law_1.2,0.10484479665756226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,1,power_law_1.2,0.11832959651947021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,1,power_law_1.2,0.15045119524002076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,1,power_law_1.2,0.11418880224227905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,1,power_law_1.2,0.12926080226898193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,1,power_law_1.2,0.16846079826354982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,1,power_law_1.2,0.26143999099731446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,1,power_law_1.2,0.3054656028747559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,1,power_law_1.2,0.3939840078353882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,1,power_law_1.2,0.20629119873046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.2,0.2658368110656738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.2,0.37547519207000735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.2,0.49266557693481444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.2,0.7060351848602295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.2,0.9299455642700195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.2,1.3616576194763184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.2,1.7974271774291992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.2,2.672831916809082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.2,3.5428863525390626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.2,4.414361572265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,1,power_law_1.2,7.028486633300782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,1,power_law_1.2,14.02568359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,1,balanced,0.2143626610438029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,1,balanced,0.3272106647491455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,1,balanced,0.5560746590296427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,1,balanced,1.0195679664611816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,1,balanced,0.1426400045553843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,1,balanced,1.8570879300435383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,1,balanced,0.21386667092641196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,1,balanced,3.495877265930176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,1,balanced,0.3335253397623698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,1,balanced,3.491413434346517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,1,balanced,0.5879253149032593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,1,balanced,3.4821065266927085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,1,balanced,0.15971733132998148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,1,balanced,1.0208640098571777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,1,balanced,3.6272265116373696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,1,balanced,0.26104533672332764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,1,balanced,1.8531200091044109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,1,balanced,3.684650739034017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,1,balanced,0.3746346632639567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,1,balanced,1.8515839576721191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,1,balanced,0.5795040130615234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,1,balanced,3.6225334803263345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,1,balanced,1.8379146258036296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,1,balanced,1.01090669631958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,1,balanced,3.610351880391439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,1,balanced,1.854629357655843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,1,balanced,1.846805254618327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,1,balanced,1.8418560028076172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,1,balanced,3.6598294576009116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,1,balanced,1.8229173024495442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,1,balanced,1.8805599212646484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,1,balanced,1.7421013514200847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,1,balanced,3.573487917582194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,1,balanced,1.7557439804077148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,1,balanced,1.856277306874593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,1,balanced,3.6331361134847007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,1,balanced,1.7113653818766277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,1,balanced,1.9545653661092122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,1,balanced,3.676410675048828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,1,balanced,1.6891093254089355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,1,balanced,1.9018719991048176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,1,balanced,1.6767627398173015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,1,balanced,3.598666508992513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,1,balanced,1.9596266746520996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,1,balanced,1.7186773618062336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,1,balanced,3.632399876912435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,1,balanced,1.9732054074605305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,1,balanced,1.7689812978108723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,1,balanced,1.7726240158081055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,1,balanced,1.897706667582194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,1,balanced,3.6393760045369468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,1,balanced,1.8406133651733398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,1,balanced,1.9819307327270508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,1,balanced,1.830143928527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,1,balanced,3.704671859741211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,1,balanced,1.971882661183675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,1,balanced,1.8386294047037761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,1,balanced,3.7450453440348306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,1,balanced,1.7846132914225261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,1,balanced,2.0396960576375327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,1,balanced,1.8741386731465657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,1,balanced,3.8563413619995117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,1,balanced,2.085498650868734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,1,balanced,1.8825279871622722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,1,balanced,3.97983455657959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,1,balanced,2.2452853520711265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,1,balanced,1.985306739807129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,1,balanced,2.409519990285238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,1,balanced,2.13374392191569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,1,balanced,6.497077306111653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,1,balanced,3.6481173833211265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,1,balanced,3.8096799850463867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,1,balanced,6.0280106862386065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,1,balanced,3.696709314982096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,1,balanced,3.714303970336914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,1,balanced,8.491621017456055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,1,balanced,5.0616105397542315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,1,balanced,5.267221450805664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,1,balanced,10.850138346354166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,1,balanced,6.721253077189128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,1,balanced,6.766213099161784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,1,balanced,0.05060799916585287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,1,balanced,0.0637066662311554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,1,balanced,0.09159466624259949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,1,balanced,13.464132944742838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,1,balanced,0.14330666263898215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,1,balanced,0.24808533986409506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,1,balanced,0.4657440185546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,1,balanced,0.47892268498738605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,1,balanced,0.49087464809417725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,1,balanced,8.209264119466146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,1,balanced,0.5049333175023397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,1,balanced,0.5163360039393107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,1,balanced,0.5409760077794393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,1,balanced,0.5652480125427246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,1,balanced,0.5892106691996256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,1,balanced,0.638597329457601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,1,balanced,0.5760106643040975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,1,balanced,0.5984853506088257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,1,balanced,0.6462933222452799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,1,balanced,8.423168182373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,1,balanced,0.6286559899648031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,1,balanced,0.6802399953206381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,1,balanced,0.6865920225779215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,1,balanced,0.7418239911397299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,1,balanced,21.400543212890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,1,balanced,1.661802609761556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,1,balanced,12.961418151855469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,1,balanced,1.9153812726338704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,1,balanced,1.2687893708546956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,1,balanced,1.3502720197041829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,1,balanced,13.403813680013021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,1,balanced,2.370783964792887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,1,balanced,2.440826733907064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,1,balanced,43.099609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,1,balanced,25.62872568766276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,1,balanced,3.4912427266438804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,1,balanced,28.238789876302082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,1,balanced,4.648458798726399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,1,balanced,9.173754374186197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,1,power_law_1.01,0.211897611618042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,1,power_law_1.01,0.31953279972076415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,1,power_law_1.01,0.3678208112716675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,1,power_law_1.01,0.7657087802886963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,1,power_law_1.01,1.2553728103637696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,1,power_law_1.01,1.6923200607299804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,1,power_law_1.01,2.456934356689453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,1,power_law_1.01,2.610495948791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,1,power_law_1.01,2.6769344329833986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,1,power_law_1.01,2.7659263610839844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,1,power_law_1.01,0.13628799915313722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,1,power_law_1.01,2.9085504531860353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,1,power_law_1.01,0.2099839925765991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,1,power_law_1.01,3.0026111602783203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,1,power_law_1.01,0.21973121166229248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,1,power_law_1.01,0.4524415969848633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,1,power_law_1.01,2.984671974182129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,1,power_law_1.01,0.6441408157348633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,1,power_law_1.01,3.0329471588134767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,1,power_law_1.01,0.955942440032959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,1,power_law_1.01,3.108550453186035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,1,power_law_1.01,1.304032039642334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,1,power_law_1.01,3.2256702423095702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,1,power_law_1.01,1.35479679107666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,1,power_law_1.01,3.282566452026367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,1,power_law_1.01,1.4064640045166015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,1,power_law_1.01,3.3571006774902346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,1,power_law_1.01,1.498521614074707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,1,power_law_1.01,3.6124671936035155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,1,power_law_1.01,1.5244352340698242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,1,power_law_1.01,3.9689537048339845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,1,power_law_1.01,1.5659968376159668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,1,power_law_1.01,4.368000030517578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,1,power_law_1.01,0.16262400150299072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,1,power_law_1.01,1.6267776489257812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,1,power_law_1.01,5.174092864990234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,1,power_law_1.01,0.258950400352478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,1,power_law_1.01,1.6697664260864258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,1,power_law_1.01,5.760537719726562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,1,power_law_1.01,0.24113919734954833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,1,power_law_1.01,1.684921646118164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,1,power_law_1.01,7.2120002746582035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,1,power_law_1.01,0.5000256061553955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,1,power_law_1.01,1.656870460510254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,1,power_law_1.01,0.6925695896148681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,1,power_law_1.01,1.695974349975586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,1,power_law_1.01,8.588377380371094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,1,power_law_1.01,0.9551360130310058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,1,power_law_1.01,1.8388799667358398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,1,power_law_1.01,11.492460632324219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,1,power_law_1.01,1.3249855995178224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,1,power_law_1.01,1.984351921081543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,1,power_law_1.01,1.3462080001831054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,1,power_law_1.01,2.1950016021728516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,1,power_law_1.01,14.179609680175782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,1,power_law_1.01,1.4445887565612794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,1,power_law_1.01,2.4259199142456054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,1,power_law_1.01,1.442412757873535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,1,power_law_1.01,17.04462127685547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,1,power_law_1.01,2.808870315551758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,1,power_law_1.01,1.470188808441162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,1,power_law_1.01,3.293824005126953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,1,power_law_1.01,1.5072319984436036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,1,power_law_1.01,25.39697265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,1,power_law_1.01,4.155904006958008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,1,power_law_1.01,1.5261119842529296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,1,power_law_1.01,4.991769790649414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,1,power_law_1.01,1.61529598236084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,1,power_law_1.01,1.616966438293457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,1,power_law_1.01,6.7320701599121096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,1,power_law_1.01,48.04446716308594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,1,power_law_1.01,1.6388288497924806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,1,power_law_1.01,8.342195129394531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,1,power_law_1.01,1.6749696731567383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,1,power_law_1.01,9.948307037353516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,1,power_law_1.01,1.7657087326049805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.01,1.8902912139892578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,1,power_law_1.01,15.026226806640626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.01,2.0845056533813477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.01,2.2774784088134767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.01,2.7295360565185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,1,power_law_1.01,29.489291381835937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.01,3.132102394104004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.01,4.0227008819580075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.01,4.83691520690918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,1,power_law_1.01,0.04857600033283234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.01,6.463565063476563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,1,power_law_1.01,0.07452160120010376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.01,8.062322998046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,1,power_law_1.01,0.08558080196380616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,1,power_law_1.01,0.1331071972846985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.01,9.685113525390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,1,power_law_1.01,0.24072959423065185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,1,power_law_1.01,0.3771136045455933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,1,power_law_1.01,14.48706512451172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,1,power_law_1.01,0.5023231983184815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,1,power_law_1.01,0.5785344123840332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,1,power_law_1.01,0.5547840118408203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,1,power_law_1.01,27.26238098144531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,1,power_law_1.01,0.639359998703003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,1,power_law_1.01,0.6711552143096924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,1,power_law_1.01,0.6868351936340332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,1,power_law_1.01,0.7056191921234131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,1,power_law_1.01,0.7918975830078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,1,power_law_1.01,0.753446388244629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,1,power_law_1.01,0.7773119926452636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,1,power_law_1.01,0.8455039978027343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,1,power_law_1.01,0.8371775627136231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.01,0.8849727630615234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.01,0.9036543846130372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.01,1.0394047737121581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.01,2.4607744216918945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.01,2.86562557220459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.01,1.6945472717285157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.01,1.957516860961914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.01,2.5190975189208986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.01,3.0926336288452148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.01,3.6552318572998046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,1,power_law_1.01,5.244876861572266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,1,power_law_1.01,9.851769256591798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,1,power_law_1.2,0.20554239749908448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,1,power_law_1.2,0.3166656017303467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,1,power_law_1.2,0.3828671932220459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,1,power_law_1.2,0.7276671886444092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,1,power_law_1.2,1.0593664169311523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,1,power_law_1.2,1.5238975524902343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,1,power_law_1.2,2.4326335906982424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,1,power_law_1.2,0.13854080438613892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,1,power_law_1.2,2.5176639556884766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,1,power_law_1.2,0.1934399962425232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,1,power_law_1.2,2.633932876586914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,1,power_law_1.2,0.21955199241638185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,1,power_law_1.2,2.7278463363647463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,1,power_law_1.2,0.38927359580993653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,1,power_law_1.2,2.883590316772461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,1,power_law_1.2,0.6090432167053222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,1,power_law_1.2,2.9900863647460936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,1,power_law_1.2,0.8379712104797363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,1,power_law_1.2,2.9808256149291994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,1,power_law_1.2,1.3012991905212403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,1,power_law_1.2,3.129734420776367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,1,power_law_1.2,1.338912010192871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,1,power_law_1.2,3.154252815246582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,1,power_law_1.2,1.4401087760925293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,1,power_law_1.2,1.4734335899353028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,1,power_law_1.2,3.3608383178710937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,1,power_law_1.2,1.5274944305419922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,1,power_law_1.2,3.4745792388916015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,1,power_law_1.2,1.5780159950256347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,1,power_law_1.2,3.688671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,1,power_law_1.2,0.1564736008644104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,1,power_law_1.2,1.5866751670837402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,1,power_law_1.2,3.836966323852539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,1,power_law_1.2,0.21413118839263917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,1,power_law_1.2,1.680134391784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,1,power_law_1.2,0.2753535985946655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,1,power_law_1.2,4.164479827880859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,1,power_law_1.2,1.72491512298584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,1,power_law_1.2,0.4478015899658203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,1,power_law_1.2,4.560428619384766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,1,power_law_1.2,1.8081535339355468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,1,power_law_1.2,0.6355199813842773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,1,power_law_1.2,5.431084823608399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,1,power_law_1.2,1.7883264541625976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,1,power_law_1.2,0.841267204284668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,1,power_law_1.2,1.989151954650879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,1,power_law_1.2,6.018521499633789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,1,power_law_1.2,1.2643391609191894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,1,power_law_1.2,1.3310272216796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,1,power_law_1.2,2.084339141845703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,1,power_law_1.2,7.443897247314453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,1,power_law_1.2,1.3297151565551757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,1,power_law_1.2,2.343814468383789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,1,power_law_1.2,8.817472076416015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,1,power_law_1.2,1.3744895935058594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,1,power_law_1.2,2.529702377319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,1,power_law_1.2,1.5070207595825196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,1,power_law_1.2,2.979385566711426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,1,power_law_1.2,11.49420166015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,1,power_law_1.2,1.5247743606567383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,1,power_law_1.2,3.418668746948242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,1,power_law_1.2,14.280581665039062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,1,power_law_1.2,1.6042688369750977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,1,power_law_1.2,4.250694274902344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,1,power_law_1.2,1.647270393371582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,1,power_law_1.2,5.095846557617188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,1,power_law_1.2,17.073504638671874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,1,power_law_1.2,1.6784128189086913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,1,power_law_1.2,6.692908477783203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,1,power_law_1.2,1.6602943420410157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,1,power_law_1.2,25.578034973144533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,1,power_law_1.2,1.7748735427856446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,1,power_law_1.2,8.373350524902344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,1,power_law_1.2,1.8938495635986328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,1,power_law_1.2,9.9614013671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.2,1.9995328903198242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.2,2.249235153198242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,1,power_law_1.2,48.44331665039063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,1,power_law_1.2,15.08417205810547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.2,2.420627212524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,1,power_law_1.2,0.048307201266288756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,1,power_law_1.2,0.07269759774208069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.2,2.867359924316406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,1,power_law_1.2,29.045803833007813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,1,power_law_1.2,0.07468799948692321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.2,3.2689727783203124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,1,power_law_1.2,0.1372159957885742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.2,4.043577575683594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,1,power_law_1.2,0.22081921100616456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.2,4.8964286804199215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,1,power_law_1.2,0.3419840097427368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.2,6.514272308349609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,1,power_law_1.2,0.4916031837463379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,1,power_law_1.2,0.5692863941192627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.2,8.130400085449219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,1,power_law_1.2,0.6204927921295166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,1,power_law_1.2,0.6072896003723145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.2,9.693510437011719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,1,power_law_1.2,0.6696512222290039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,1,power_law_1.2,14.345645141601562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,1,power_law_1.2,0.6896512031555175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,1,power_law_1.2,0.7196479797363281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,1,power_law_1.2,0.8011520385742188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,1,power_law_1.2,26.859774780273437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,1,power_law_1.2,0.7722367763519287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,1,power_law_1.2,0.8009344100952148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,1,power_law_1.2,0.9175807952880859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,1,power_law_1.2,0.8506431579589844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.2,0.9307519912719726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.2,0.9774975776672363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.2,1.0754688262939454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.2,2.5650047302246093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.2,2.964121627807617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.2,1.7981887817382813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.2,2.0311424255371096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.2,2.5665279388427735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.2,3.1581632614135744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.2,3.6630462646484374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,1,power_law_1.2,5.395103836059571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,1,power_law_1.2,9.8897216796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,1,balanced,0.045968001087506614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,1,balanced,0.05482133229573568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,1,balanced,0.07101333141326904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,1,balanced,0.052229334910710655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,1,balanced,0.10578133662541707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,1,balanced,0.0629066675901413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,1,balanced,0.16226133704185486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,1,balanced,0.09268800417582194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,1,balanced,0.14801599582036337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,1,balanced,0.14779200156529745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,1,balanced,0.24897066752115884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,1,balanced,0.14525866508483887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,1,balanced,0.2453226645787557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,1,balanced,0.14985600113868713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,1,balanced,0.23919999599456787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,1,balanced,0.14967466394106546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,1,balanced,0.25256532430648804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,1,balanced,0.1543839971224467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,1,balanced,0.24200532833735147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,1,balanced,0.15832533439000449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,1,balanced,0.24966400861740112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,1,balanced,0.15682133038838705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,1,balanced,0.23811199267705283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,1,balanced,0.16425599654515585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,1,balanced,0.2579893271128337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,1,balanced,0.16153599818547568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,1,balanced,0.2546186645825704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,1,balanced,0.16929600636164346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,1,balanced,0.2512906591097514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,1,balanced,0.17566933234532675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,1,balanced,0.25833600759506226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,1,balanced,0.05853333572546641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,1,balanced,0.06968000034491222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,1,balanced,0.18074132998784384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,1,balanced,0.25621867179870605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,1,balanced,0.08569066723187764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,1,balanced,0.26100800434748334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,1,balanced,0.12878400087356567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,1,balanced,0.2062293291091919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,1,balanced,0.18492267529169717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,1,balanced,0.2748746673266093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,1,balanced,0.18207999070485434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,1,balanced,0.22340265909830728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,1,balanced,0.17960532506306967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,1,balanced,0.28726933399836224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,1,balanced,0.17062934239705405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,1,balanced,0.2682186762491862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,1,balanced,0.16993065675099692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,1,balanced,0.17141334215799967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,1,balanced,0.3171413342158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,1,balanced,0.30667734146118164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,1,balanced,0.17825599511464438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,1,balanced,0.18195199966430664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,1,balanced,0.34486933549245197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,1,balanced,0.1746293306350708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,1,balanced,0.18604799111684164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,1,balanced,0.4366613229115804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,1,balanced,0.19250667095184326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,1,balanced,0.47804800669352215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,1,balanced,0.20107734203338623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,1,balanced,0.19960532585779825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,1,balanced,0.5212213198343912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,1,balanced,0.22492265701293945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,1,balanced,0.5191466808319092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,1,balanced,0.23937066396077475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,1,balanced,0.01941866676012675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,1,balanced,0.28806400299072266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,1,balanced,0.02447466552257538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,1,balanced,0.7323893706003824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,1,balanced,0.026000000536441803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,1,balanced,0.7054826418558756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,1,balanced,0.030271999537944794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,1,balanced,0.3283039927482605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,1,balanced,0.048725331823031105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,1,balanced,0.049546668926874794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,1,balanced,0.05070933202902476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,1,balanced,0.5030239820480347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,1,balanced,0.05266666909058889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,1,balanced,0.05529599885145823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,1,balanced,0.9574399789174398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,1,balanced,0.056618665655454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,1,balanced,0.9011893272399902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,1,balanced,0.060906668504079185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,1,balanced,0.058864002426465355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,1,balanced,0.5932000080744425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,1,balanced,0.06033066908518473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,1,balanced,0.06358399987220764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,1,balanced,0.06286400059858958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,1,balanced,0.0639626681804657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,1,balanced,0.06886933247248332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,1,balanced,0.8541279633839926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,1,balanced,1.397605260213216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,1,balanced,0.07519466678301494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,1,balanced,1.3147679964701335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,1,balanced,0.07910933097203572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,1,balanced,0.1297706663608551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,1,balanced,1.1091679732004802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,1,balanced,0.15110400319099426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,1,balanced,0.14036267002423605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,1,balanced,1.8487040201822917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,1,balanced,1.759130636850993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,1,balanced,0.140773336092631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,1,balanced,1.610640048980713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,1,balanced,0.2526986598968506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,1,balanced,2.2868800163269043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,1,balanced,2.0419519742329917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,1,balanced,2.171109358469645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,1,balanced,0.2408533294995626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,1,balanced,0.34459733963012695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,1,balanced,2.4909440676371255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,1,balanced,3.5729974110921225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,1,balanced,3.420149485270182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,1,balanced,0.44764800866444904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,1,balanced,3.84718386332194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,1,balanced,0.5500373442967733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,1,balanced,6.9417069753011065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,1,balanced,0.865226666132609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,1,balanced,6.782426834106445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,1,balanced,7.4476744333903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,1,balanced,1.6895947456359863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,1,power_law_1.01,0.040678399801254275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,1,power_law_1.01,0.046988800168037415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,1,power_law_1.01,0.055801600217819214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,1,power_law_1.01,0.04806399941444397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,1,power_law_1.01,0.07754240036010743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,1,power_law_1.01,0.051507198810577394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,1,power_law_1.01,0.09194880127906799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,1,power_law_1.01,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,1,power_law_1.01,0.1312831997871399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,1,power_law_1.01,0.10671360492706299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,1,power_law_1.01,0.13738240003585817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,1,power_law_1.01,0.13722239732742308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,1,power_law_1.01,0.1415552020072937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,1,power_law_1.01,0.20487680435180664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,1,power_law_1.01,0.14477440118789672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,1,power_law_1.01,0.14487040042877197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,1,power_law_1.01,0.21420800685882568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,1,power_law_1.01,0.15119999647140503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,1,power_law_1.01,0.22216320037841797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,1,power_law_1.01,0.15288959741592406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,1,power_law_1.01,0.22099199295043945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,1,power_law_1.01,0.15548800230026244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,1,power_law_1.01,0.15985920429229736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,1,power_law_1.01,0.16210559606552125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,1,power_law_1.01,0.1700160026550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,1,power_law_1.01,0.1866688013076782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,1,power_law_1.01,0.21182079315185548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,1,power_law_1.01,0.24161279201507568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,1,power_law_1.01,0.05653759837150574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,1,power_law_1.01,0.29592959880828856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,1,power_law_1.01,0.06285439729690552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,1,power_law_1.01,0.3416192054748535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,1,power_law_1.01,0.07760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,1,power_law_1.01,0.45944957733154296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,1,power_law_1.01,0.1019327998161316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,1,power_law_1.01,0.564409589767456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,1,power_law_1.01,0.11703679561614991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,1,power_law_1.01,0.7738239765167236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,1,power_law_1.01,0.15745279788970948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,1,power_law_1.01,0.989510440826416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,1,power_law_1.01,1.4192831993103028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,1,power_law_1.01,1.8343551635742188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,1,power_law_1.01,2.2471744537353517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,1,power_law_1.01,3.5214080810546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,1,power_law_1.01,6.915526580810547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,1,power_law_1.01,0.01886080056428909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,1,power_law_1.01,0.021132799983024596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,1,power_law_1.01,0.02479359954595566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,1,power_law_1.01,0.030585598945617676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,1,power_law_1.01,0.039084801077842714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,1,power_law_1.01,0.05398399829864502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,1,power_law_1.01,0.05729280114173889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,1,power_law_1.01,0.0592960000038147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,1,power_law_1.01,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,1,power_law_1.01,0.06473600268363952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,1,power_law_1.01,0.06948480010032654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,1,power_law_1.01,0.06537600159645081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,1,power_law_1.01,0.06736000180244446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,1,power_law_1.01,0.07098879814147949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,1,power_law_1.01,0.07148799896240235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,1,power_law_1.01,0.07208960056304932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,1,power_law_1.01,0.0753600001335144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,1,power_law_1.01,0.08358399868011475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,1,power_law_1.01,0.09201920032501221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,1,power_law_1.01,0.17613439559936522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,1,power_law_1.01,0.19876480102539062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,1,power_law_1.01,0.16135040521621705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,1,power_law_1.01,0.17799680233001708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,1,power_law_1.01,0.23423359394073487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,1,power_law_1.01,0.2839807987213135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,1,power_law_1.01,0.3896512031555176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,1,power_law_1.01,0.4899903774261475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,1,power_law_1.01,0.5875455856323242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,1,power_law_1.01,0.8979840278625488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,1,power_law_1.01,1.7361536026000977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,1,power_law_1.2,0.04890879988670349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,1,power_law_1.2,0.05273600220680237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,1,power_law_1.2,0.07316480278968811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,1,power_law_1.2,0.10161919593811035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,1,power_law_1.01,0.23059840202331544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,1,power_law_1.2,0.13315839767456056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,1,power_law_1.01,0.23710079193115235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,1,power_law_1.2,0.20641920566558838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,1,power_law_1.01,0.24034559726715088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,1,power_law_1.2,0.22264320850372316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,1,power_law_1.01,0.2477247953414917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,1,power_law_1.2,0.23148798942565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,1,power_law_1.01,0.24327681064605713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,1,power_law_1.2,0.23739519119262695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,1,power_law_1.01,0.24822399616241456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,1,power_law_1.2,0.24240000247955323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,1,power_law_1.01,0.2619328022003174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,1,power_law_1.2,0.2362112045288086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,1,power_law_1.01,0.26625919342041016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,1,power_law_1.2,0.24186880588531495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,1,power_law_1.01,0.29840641021728515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,1,power_law_1.2,0.2522880077362061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,1,power_law_1.01,0.31731200218200684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,1,power_law_1.2,0.24735360145568847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,1,power_law_1.01,0.3752063989639282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,1,power_law_1.2,0.25601921081542967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,1,power_law_1.2,0.2656703948974609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,1,power_law_1.01,0.4204415798187256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,1,power_law_1.01,0.15926400423049927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,1,power_law_1.2,0.2732736110687256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,1,power_law_1.01,0.5402175903320312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,1,power_law_1.01,0.16206079721450806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,1,power_law_1.2,0.3019711971282959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,1,power_law_1.01,0.6343488216400146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,1,power_law_1.01,0.17115520238876342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,1,power_law_1.2,0.3273855924606323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,1,power_law_1.01,0.8487680435180665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,1,power_law_1.01,0.17380479574203492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,1,power_law_1.2,0.384768009185791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,1,power_law_1.01,1.0415231704711914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,1,power_law_1.01,0.1749824047088623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,1,power_law_1.2,0.4251391887664795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,1,power_law_1.01,1.4398143768310547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,1,power_law_1.01,0.1776576042175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,1,power_law_1.2,0.541644811630249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,1,power_law_1.01,1.844550323486328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,1,power_law_1.01,0.18039040565490722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,1,power_law_1.2,0.6279104232788086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,1,power_law_1.01,2.2482751846313476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,1,power_law_1.01,0.18407679796218873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,1,power_law_1.2,0.8410816192626953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,1,power_law_1.01,3.4602497100830076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,1,power_law_1.01,0.1925055980682373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,1,power_law_1.2,1.0340160369873046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,1,power_law_1.01,0.19549440145492553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,1,power_law_1.2,1.4235648155212401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,1,power_law_1.01,6.6953857421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,1,power_law_1.01,0.21276159286499025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,1,power_law_1.2,1.8216512680053711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,1,power_law_1.01,0.23927679061889648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,1,power_law_1.2,2.2152767181396484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,1,power_law_1.01,0.26946558952331545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,1,power_law_1.01,0.3312000036239624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,1,power_law_1.2,3.4243328094482424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,1,power_law_1.01,0.39123198986053465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,1,power_law_1.2,6.675513458251953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,1,power_law_1.01,0.5101632118225098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,1,power_law_1.01,0.6341824054718017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,1,power_law_1.01,0.8632767677307129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,1,power_law_1.01,1.0971455574035645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,1,power_law_1.01,1.55282564163208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,1,power_law_1.01,2.0061439514160155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,1,power_law_1.01,2.461478424072266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,1,power_law_1.01,3.81014404296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,1,power_law_1.2,0.04330880045890808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,1,power_law_1.01,7.414694213867188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,1,power_law_1.2,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,1,power_law_1.2,0.05490559935569763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,1,power_law_1.2,0.07196800112724304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,1,power_law_1.2,0.08982399702072144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,1,power_law_1.2,0.13098880052566528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,1,power_law_1.2,0.1377087950706482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,1,power_law_1.2,0.14318079948425294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,1,power_law_1.2,0.1454208016395569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,1,power_law_1.2,0.14648319482803346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,1,power_law_1.2,0.1540735960006714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,1,power_law_1.2,0.15571199655532836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,1,power_law_1.2,0.15771520137786865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,1,power_law_1.2,0.1619968056678772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,1,power_law_1.2,0.17112319469451903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,1,power_law_1.2,0.174508798122406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,1,power_law_1.2,0.18828799724578857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,1,power_law_1.2,0.21700479984283447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,1,power_law_1.2,0.24159998893737794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,1,power_law_1.2,0.29628159999847414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,1,power_law_1.2,0.3447551965713501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,1,power_law_1.2,0.45866241455078127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,1,power_law_1.2,0.56430082321167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,1,power_law_1.2,0.774022388458252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,1,power_law_1.2,0.9884799957275391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,1,power_law_1.2,1.4034687995910644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,1,power_law_1.2,1.8155071258544921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,1,power_law_1.2,2.2478271484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,1,power_law_1.2,3.505855941772461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,1,power_law_1.2,6.876614379882812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,1,power_law_1.2,0.018464000523090364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,1,power_law_1.2,0.02172800004482269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,1,power_law_1.2,0.024806399643421174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,1,power_law_1.2,0.05616000294685364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,1,power_law_1.2,0.028896000981330872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,1,power_law_1.2,0.061222398281097413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,1,power_law_1.2,0.03925760090351105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,1,power_law_1.2,0.07252479791641235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,1,power_law_1.2,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,1,power_law_1.2,0.0849407970905304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,1,power_law_1.2,0.058406400680541995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,1,power_law_1.2,0.10660480260848999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,1,power_law_1.2,0.06110720038414001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,1,power_law_1.2,0.16327040195465087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,1,power_law_1.2,0.06230400204658508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,1,power_law_1.2,0.16551040410995482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,1,power_law_1.2,0.0663424015045166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,1,power_law_1.2,0.1690816044807434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,1,power_law_1.2,0.07094399929046631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,1,power_law_1.2,0.16838400363922118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,1,power_law_1.2,0.06424319744110107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,1,power_law_1.2,0.17359999418258668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,1,power_law_1.2,0.06968960165977478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,1,power_law_1.2,0.18174079656600953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,1,power_law_1.2,0.07222399711608887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,1,balanced,0.10756267110506694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,1,balanced,0.1696959932645162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,1,power_law_1.2,0.18506239652633666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,1,power_law_1.2,0.0707264006137848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,1,balanced,0.27138666311899823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,1,power_law_1.2,0.17946879863739013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,1,power_law_1.2,0.07477759718894958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,1,balanced,0.47704001267751056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,1,power_law_1.2,0.18328959941864015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,1,power_law_1.2,0.08144000172615051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,1,balanced,0.8560106754302979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,1,power_law_1.2,0.19409919977188111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,1,power_law_1.2,0.08419839739799499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,1,balanced,0.8468426863352457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,1,balanced,0.8512852986653646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,1,power_law_1.2,0.20184319019317626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,1,power_law_1.2,0.09336320161819459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,1,balanced,0.8535786469777426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,1,power_law_1.2,0.21377279758453369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,1,power_law_1.2,0.17585279941558837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,1,balanced,0.8638239701588949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,1,power_law_1.2,0.24578559398651123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,1,power_law_1.2,0.20629758834838868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,1,balanced,0.8449493249257406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,1,power_law_1.2,0.2732608079910278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,1,power_law_1.2,0.168121600151062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,1,balanced,0.8487573464711508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,1,power_law_1.2,0.3313600063323975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,1,power_law_1.2,0.18929920196533204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,1,balanced,0.8772479693094889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,1,power_law_1.2,0.39134719371795657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,1,power_law_1.2,0.2401279926300049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,1,balanced,0.8602186838785807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,1,power_law_1.2,0.5101183891296387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,1,power_law_1.2,0.294323205947876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,1,balanced,0.8577919801076254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,1,power_law_1.2,0.6283199787139893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,1,balanced,0.8529439767201742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,1,power_law_1.2,0.3901439905166626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,1,power_law_1.2,0.8615872383117675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,1,balanced,0.8687626520792643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,1,power_law_1.2,0.4982463836669922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,1,balanced,0.8533013661702474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,1,power_law_1.2,1.089798355102539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,1,power_law_1.2,0.5921599864959717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,1,power_law_1.2,1.541651153564453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,1,power_law_1.2,0.9072256088256836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,1,balanced,0.8631573518117269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,1,power_law_1.2,1.9884672164916992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,1,power_law_1.2,1.7404415130615234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,1,balanced,0.8631359736124674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,1,power_law_1.2,2.4444864273071287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,1,balanced,0.9098133246103922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,1,power_law_1.2,3.7907646179199217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,1,balanced,0.9506826400756836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,1,power_law_1.2,7.41751708984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,1,balanced,1.4508372942606609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,1,balanced,1.3858240445454915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,1,balanced,0.07169066866238911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,1,balanced,1.8830506006876628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,1,balanced,0.11037866274515788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,1,balanced,0.1681119998296102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,1,balanced,0.2677226662635803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,1,balanced,0.4723786513010661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,1,balanced,0.4412533442179362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,1,balanced,0.45713067054748535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,1,balanced,0.48926401138305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,1,balanced,2.4590880076090493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,1,balanced,0.44223999977111816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,1,balanced,0.47898133595784503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,1,balanced,0.4542613426844279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,1,balanced,0.4693066676457723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,1,balanced,0.49180801709493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,1,balanced,0.45581865310668945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,1,balanced,0.46952001253763836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,1,balanced,0.47173865636189777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,1,balanced,3.6582508087158203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,1,balanced,0.48998932043711346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,1,balanced,0.5057173172632853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,1,balanced,0.5290666818618774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,1,balanced,0.5874933401743571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,1,balanced,0.6415573358535767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,1,balanced,4.772874514261882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,1,balanced,0.9342346986134847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,1,balanced,0.9881386756896973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,1,balanced,1.3705546061197917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,1,balanced,6.012837092081706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,1,balanced,1.781541347503662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,1,balanced,2.6293439865112305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,1,balanced,9.80407460530599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,1,balanced,3.463754653930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,1,balanced,0.08907199899355571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,1,balanced,0.13038399815559387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,1,balanced,4.298581441243489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,1,balanced,0.20100800196329752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,1,balanced,0.31620800495147705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,1,balanced,0.5025279919306437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,1,balanced,0.45579731464385986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,1,balanced,0.4527839819590251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,1,balanced,0.4608000119527181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,1,balanced,0.4563680092493693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,1,balanced,0.46559464931488037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,1,balanced,0.028714666763941448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,1,balanced,0.48898132642110187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,1,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,1,balanced,0.4554400046666463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,1,balanced,0.05218133330345154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,1,balanced,0.48773332436879474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,1,balanced,0.07328533132870992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,1,balanced,0.11742933591206868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,1,balanced,0.4504266579945882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,1,balanced,0.12090667088826497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,1,balanced,0.4666186571121216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,1,balanced,0.1267626682917277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,1,balanced,0.4825119972229004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,1,balanced,0.13411200046539307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,1,balanced,0.1397760013739268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,1,balanced,0.48051198323567706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,1,balanced,0.14527466893196106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,1,balanced,0.4926079909006755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,1,balanced,0.15781333049138388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,1,balanced,0.14282133181889853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,1,balanced,0.5213599999745687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,1,balanced,0.14923200011253357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,1,balanced,6.792661031087239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,1,balanced,0.16080000003178915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,1,balanced,0.5629493395487467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,1,balanced,0.15504533052444458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,1,balanced,0.1576586663722992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,1,balanced,0.6217013200124105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,1,balanced,0.17128000656763712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,1,balanced,0.18764267365137735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,1,balanced,20.417199452718098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,1,balanced,1.0194239616394043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,1,balanced,0.19089599450429282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,1,balanced,0.3916586637496948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,1,balanced,1.0953599611918132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,1,balanced,0.4530986547470093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,1,balanced,0.3347253402074178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,1,balanced,1.536784013112386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,1,balanced,0.3312480052312215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,1,balanced,0.6018720070521036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,1,balanced,1.8800746599833171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,1,balanced,0.5767306486765543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,1,balanced,2.7324905395507812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,1,balanced,13.788832346598307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,1,balanced,0.8319466908772787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,1,balanced,3.6159092585245767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,1,balanced,1.0867253144582112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,1,balanced,4.449695905049642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,1,balanced,1.3495519955952961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,1,balanced,7.010170618693034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,1,balanced,2.1254773139953613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,1,power_law_1.01,0.10547200441360474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,1,power_law_1.01,0.1161344051361084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,1,power_law_1.01,0.2122431993484497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,1,power_law_1.01,0.33251199722290037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,1,power_law_1.01,0.43131518363952637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,1,power_law_1.01,0.7001279830932617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,1,power_law_1.01,0.7401279926300048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,1,power_law_1.01,0.7761343955993653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,1,power_law_1.01,0.7796735763549805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,1,power_law_1.01,0.8033408164978028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,1,power_law_1.01,0.8180288314819336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,1,power_law_1.01,0.8144576072692871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,1,power_law_1.01,0.8328512191772461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,1,power_law_1.01,0.851097583770752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,1,power_law_1.01,0.8548800468444824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,1,power_law_1.01,0.8559935569763184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,1,power_law_1.01,0.8865471839904785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,1,power_law_1.01,0.9680576324462891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,1,power_law_1.01,1.041107177734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,1,power_law_1.01,1.1995264053344727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,1,power_law_1.01,1.345580768585205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,1,power_law_1.01,1.686739158630371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,1,balanced,4.20249080657959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,1,power_law_1.01,1.9589056015014648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,1,balanced,13.750170389811197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,1,power_law_1.01,2.5656448364257813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,1,power_law_1.01,3.16549129486084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,1,power_law_1.01,0.07707520127296448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,1,power_law_1.01,0.07638400197029113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,1,power_law_1.01,4.295622253417969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,1,power_law_1.01,0.12643840312957763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,1,power_law_1.01,5.426566314697266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,1,power_law_1.01,0.1852288007736206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,1,power_law_1.01,6.659846496582031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,1,power_law_1.01,0.24766080379486083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,1,power_law_1.01,0.3878528118133545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,1,power_law_1.01,10.251296234130859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,1,power_law_1.01,0.4106751918792725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,1,power_law_1.01,0.4250944137573242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,1,power_law_1.01,20.636090087890626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,1,power_law_1.01,0.423967981338501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,1,power_law_1.01,0.4337791919708252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,1,power_law_1.01,0.43079681396484376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,1,power_law_1.01,0.4501632213592529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,1,power_law_1.01,0.46643838882446287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,1,power_law_1.01,0.45045762062072753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,1,power_law_1.01,0.47358078956604005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,1,power_law_1.01,0.4864511966705322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,1,power_law_1.01,0.503001594543457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,1,power_law_1.01,0.5644544124603271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,1,power_law_1.01,0.6060800075531005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,1,power_law_1.01,0.7296512126922607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,1,power_law_1.01,0.8134783744812012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,1,power_law_1.01,1.0425408363342286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,1,power_law_1.01,1.2434304237365723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,1,power_law_1.01,1.672447967529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,1,power_law_1.01,2.056268882751465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,1,power_law_1.01,2.8672704696655273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,1,power_law_1.01,3.6790847778320312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,1,power_law_1.01,4.496384048461914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,1,power_law_1.01,6.954624176025391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,1,power_law_1.01,13.580767822265624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,1,power_law_1.01,0.026182401180267333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,1,power_law_1.01,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,1,power_law_1.01,0.04650880098342895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,1,power_law_1.01,0.07241600155830383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,1,power_law_1.01,0.10184320211410522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,1,power_law_1.01,0.1487936019897461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,1,power_law_1.01,0.08549759984016418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,1,power_law_1.01,0.16821119785308838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,1,power_law_1.01,0.09026560187339783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,1,power_law_1.01,0.1664512038230896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,1,power_law_1.01,0.13063039779663085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,1,power_law_1.01,0.20039041042327882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,1,power_law_1.01,0.17323520183563232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,1,power_law_1.01,0.2704063892364502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,1,power_law_1.01,0.1814463973045349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,1,power_law_1.01,0.3994879961013794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,1,power_law_1.01,0.18480000495910645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,1,power_law_1.2,0.09896960258483886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,1,power_law_1.01,0.4175871849060059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,1,power_law_1.01,0.18714239597320556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,1,power_law_1.2,0.12202240228652954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,1,power_law_1.01,0.45026559829711915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,1,power_law_1.01,0.19497599601745605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,1,power_law_1.01,0.4348480224609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,1,power_law_1.2,0.2055743932723999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,1,power_law_1.01,0.20690560340881348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,1,power_law_1.01,0.45431041717529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,1,power_law_1.01,0.19803520441055297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,1,power_law_1.2,0.2901760101318359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,1,power_law_1.01,0.46305279731750487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,1,power_law_1.01,0.20440320968627929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,1,power_law_1.2,0.3903167963027954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,1,power_law_1.01,0.4658368110656738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,1,power_law_1.01,0.21400320529937744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,1,power_law_1.2,0.7214784145355224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,1,power_law_1.01,0.47675518989562987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,1,power_law_1.01,0.23425920009613038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,1,power_law_1.2,0.7819456100463867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,1,power_law_1.01,0.4707647800445557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.01,0.25454719066619874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,1,power_law_1.2,0.7960319995880127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,1,power_law_1.01,0.4758463859558105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.01,0.5834176063537597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,1,power_law_1.2,0.8079936027526855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,1,power_law_1.01,0.4897791862487793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.01,0.663865613937378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,1,power_law_1.2,0.8376576423645019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,1,power_law_1.01,0.5212096214294434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.01,0.41182718276977537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,1,power_law_1.2,0.82675199508667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,1,power_law_1.01,0.591923189163208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.01,0.46767358779907225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,1,power_law_1.2,0.8304191589355469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.01,0.6258560180664062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.01,0.578220796585083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,1,power_law_1.2,0.8715264320373535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.01,0.7373055934906005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.01,0.6991936206817627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.01,0.8378368377685547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,1,power_law_1.2,0.8541888236999512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.01,0.9508223533630371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.01,1.087615966796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,1,power_law_1.2,0.872332763671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.01,1.207366371154785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.01,1.2823936462402343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,1,power_law_1.2,0.8978816032409668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.01,1.4600768089294434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.01,1.7037696838378906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,1,power_law_1.2,0.9245504379272461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,1,power_law_1.01,2.2419647216796874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,1,power_law_1.2,0.07663999795913697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.01,2.1374143600463866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,1,power_law_1.2,0.9939519882202148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,1,power_law_1.2,0.08364160060882568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.01,2.963737678527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,1,power_law_1.2,1.069478416442871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,1,power_law_1.01,4.301375961303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,1,power_law_1.2,0.1194815993309021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,1,power_law_1.2,1.2478912353515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.01,3.7977985382080077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,1,power_law_1.2,0.1673535943031311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,1,power_law_1.2,1.387929630279541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,1,power_law_1.2,0.22794880867004394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.01,4.640268707275391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,1,power_law_1.2,0.3788480043411255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,1,power_law_1.2,1.6897855758666993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,1,power_law_1.01,7.085721588134765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,1,power_law_1.2,0.4170112133026123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,1,power_law_1.2,1.9700992584228516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,1,power_law_1.2,0.4315455913543701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,1,power_law_1.2,2.5700416564941406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,1,power_law_1.01,13.808224487304688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,1,power_law_1.2,0.4456575870513916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,1,power_law_1.2,3.12806396484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,1,power_law_1.2,0.4432384014129639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,1,power_law_1.2,0.45593600273132323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,1,power_law_1.2,4.2906944274902346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,1,power_law_1.2,0.4684351921081543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,1,power_law_1.2,5.405311965942383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,1,power_law_1.2,0.45925121307373046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,1,power_law_1.2,0.46631679534912107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,1,power_law_1.2,6.535756683349609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,1,power_law_1.2,0.4901951789855957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,1,power_law_1.2,10.19863052368164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,1,power_law_1.2,0.4922368049621582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,1,power_law_1.2,0.5289408206939697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,1,power_law_1.2,0.5838463783264161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,1,power_law_1.2,20.25557098388672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,1,power_law_1.2,0.6156991958618164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,1,power_law_1.2,0.7406720161437989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,1,power_law_1.2,0.8284799575805664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,1,power_law_1.2,1.0686335563659668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,1,power_law_1.2,1.2469311714172364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,1,power_law_1.2,1.6567680358886718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,1,power_law_1.2,2.036569595336914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,1,power_law_1.2,2.860281562805176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,1,power_law_1.2,3.6285057067871094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,1,power_law_1.2,4.466508865356445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,1,power_law_1.2,6.906956481933594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,1,power_law_1.2,13.403724670410156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,1,power_law_1.2,0.08295680284500122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,1,power_law_1.2,0.09720320105552674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,1,power_law_1.2,0.1254207968711853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,1,power_law_1.2,0.1966976046562195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,1,power_law_1.2,0.23866240978240966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,1,power_law_1.2,0.41614718437194825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,1,power_law_1.2,0.4444608211517334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,1,power_law_1.2,0.02654080092906952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,1,power_law_1.2,0.4558591842651367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,1,power_law_1.2,0.030771198868751525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,1,power_law_1.2,0.45676798820495607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,1,power_law_1.2,0.04854399859905243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,1,power_law_1.2,0.4643392086029053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,1,power_law_1.2,0.06618239879608154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,1,power_law_1.2,0.4721792221069336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,1,power_law_1.2,0.0906175971031189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,1,power_law_1.2,0.46300158500671384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,1,power_law_1.2,0.1540544033050537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,1,power_law_1.2,0.47930240631103516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,1,power_law_1.2,0.1617408037185669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,1,power_law_1.2,0.48664321899414065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,1,power_law_1.2,0.18042880296707153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,1,power_law_1.2,0.5001664161682129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,1,power_law_1.2,0.5054848194122314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,1,power_law_1.2,0.17511039972305298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,1,power_law_1.2,0.5241663932800293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,1,power_law_1.2,0.1783679962158203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,1,power_law_1.2,0.5852543830871582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,1,power_law_1.2,0.1969151973724365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,1,balanced,0.21399466196695963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,1,balanced,0.3582506577173869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,1,power_law_1.2,0.19162240028381347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,1,balanced,0.5933226744333903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,1,power_law_1.2,0.19896320104599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,1,balanced,1.076533317565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,1,power_law_1.2,0.20826880931854247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,1,balanced,1.9656640688578289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,1,power_law_1.2,0.19444479942321777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,1,balanced,2.4281333287556968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,1,power_law_1.2,0.2114176034927368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,1,balanced,2.403722604115804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,1,power_law_1.2,0.21731200218200683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,1,balanced,2.4702773094177246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,1,power_law_1.2,0.22645120620727538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,1,balanced,2.5787413914998374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,1,balanced,2.5081119537353516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.2,0.26851840019226075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,1,balanced,2.4482827186584473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.2,0.5805247783660888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,1,balanced,2.4907520612080893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.2,0.6816319942474365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,1,balanced,0.14711466431617737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,1,balanced,2.4983466466267905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.2,0.4012864112854004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,1,balanced,0.22458134094874063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,1,balanced,2.4919519424438477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.2,0.4718207836151123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,1,balanced,0.35868799686431885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.2,0.6013760089874267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,1,balanced,2.550704002380371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,1,balanced,0.5949653387069702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,1,balanced,1.0551199913024902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.2,0.7364096164703369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,1,balanced,2.499349276224772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,1,balanced,1.288815975189209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.2,0.9710975646972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,1,balanced,2.537567933400472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,1,balanced,1.2626773516337078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.2,1.1923839569091796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,1,balanced,1.2404692967732747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,1,balanced,2.554080009460449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,1,balanced,1.3075786431630452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.2,1.4802495956420898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,1,balanced,1.2942879994710286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,1,balanced,2.546463966369629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,1,power_law_1.2,2.250873565673828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,1,balanced,1.2756799856821697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,1,balanced,1.384127934773763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,1,balanced,2.576133410135905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,1,power_law_1.2,4.325785446166992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,1,balanced,1.3101812998453777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,1,balanced,1.337125301361084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,1,balanced,2.624885400136312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,1,balanced,1.303594668706258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,1,balanced,1.3510665893554688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,1,balanced,4.5400746663411455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,1,balanced,1.3253813584645588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,1,balanced,1.363269329071045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,1,balanced,1.3797705968221028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,1,balanced,4.103919982910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,1,balanced,1.43068265914917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,1,balanced,1.4943307240804036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,1,balanced,6.174287796020508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,1,balanced,2.512634595235189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,1,balanced,8.405248006184896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,1,balanced,2.357759952545166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,1,balanced,3.555946667989095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,1,balanced,8.998640060424805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,1,balanced,4.673493385314941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,1,balanced,13.342491149902344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,1,balanced,5.342704137166341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,1,balanced,13.80184555053711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,1,balanced,7.705381393432617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,1,balanced,8.302597045898438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,1,balanced,23.304550170898438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.2,0.641267204284668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.2,0.7639296054840088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.2,0.8400128364562989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.2,1.0696767807006835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.2,1.2796287536621094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.2,1.7051136016845703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.2,2.112531280517578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.2,2.945043182373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.2,3.735270309448242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,1,balanced,0.15552533666292825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,1,balanced,0.25457600752512616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.2,4.5759422302246096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,1,balanced,0.37149866422017414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,1,balanced,0.6079306602478027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,1,power_law_1.2,6.980684661865235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,1,balanced,1.069760004679362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,1,balanced,13.6310666402181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,1,balanced,1.2700426578521729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,1,balanced,1.2286826769510906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,1,power_law_1.2,13.643238830566407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,1,balanced,1.2154453595479329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,1,balanced,1.2653813362121582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,1,balanced,1.2326080004374187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,1,balanced,1.2152586778004963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,1,balanced,1.2869439919789631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,1,balanced,1.2610026995340984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,1,balanced,1.3599680264790852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,1,balanced,1.2459839979807537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,1,balanced,1.290778636932373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,1,balanced,1.3005013465881348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,1,balanced,1.3057280381520588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,1,balanced,1.2879040241241455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,1,balanced,1.3014933268229167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,1,balanced,48.4450429280599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,1,balanced,1.3446720441182454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,1,balanced,2.488639990488688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,1,balanced,2.4632213910420737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,1,balanced,3.569925308227539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,1,balanced,4.6062774658203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,1,balanced,27.71026611328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,1,balanced,5.348949432373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,1,balanced,7.688922882080078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,1,balanced,8.265381495157877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,1,balanced,0.05217599868774414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,1,balanced,0.06803733110427856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,1,balanced,0.09672533472379048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,1,balanced,0.14999999602635702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,1,balanced,0.26018667221069336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,1,balanced,0.32344533999760944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,1,balanced,0.33575467268625897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,1,balanced,0.3505493402481079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,1,balanced,0.3618026574452718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,1,balanced,13.57757314046224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,1,balanced,0.3766506512959798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,1,balanced,0.4031466643015544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,1,balanced,0.4289439916610718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,1,balanced,0.38532265027364093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,1,balanced,0.4094666639963786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,1,balanced,0.43480531374613446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,1,balanced,0.40295998255411786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,1,balanced,0.4289919932683309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,1,balanced,0.4952319860458374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,1,balanced,0.4753386576970418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,1,balanced,1.0172906716664631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,1,balanced,1.149290641148885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,1,balanced,0.8441119988759359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,1,power_law_1.01,0.21804161071777345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,1,balanced,0.818885326385498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,1,power_law_1.01,0.3077824115753174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,1,power_law_1.01,0.44752001762390137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,1,power_law_1.01,0.7945280075073242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,1,power_law_1.01,1.1205504417419434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,1,power_law_1.01,1.898636817932129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,1,balanced,1.4759306907653809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,1,power_law_1.01,1.994175910949707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,1,power_law_1.01,2.0951295852661134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,1,power_law_1.01,2.187379264831543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,1,power_law_1.01,2.256902313232422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,1,balanced,1.5403626759847004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,1,power_law_1.01,2.212179183959961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,1,power_law_1.01,2.230656051635742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,1,power_law_1.01,2.32174072265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,1,power_law_1.01,2.3152576446533204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,1,power_law_1.01,2.3936960220336916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,1,power_law_1.01,0.14451839923858642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,1,power_law_1.01,2.3924415588378904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,1,power_law_1.01,0.18113919496536254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,1,power_law_1.01,2.499513626098633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,1,balanced,2.283616065979004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,1,power_law_1.01,2.695712089538574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,1,balanced,27.053876241048176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,1,power_law_1.01,2.86627197265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,1,power_law_1.01,3.1901952743530275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,1,power_law_1.01,3.6224193572998047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,1,power_law_1.01,4.277369689941406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,1,power_law_1.01,4.944953536987304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,1,power_law_1.01,6.449759674072266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,1,balanced,2.936181386311849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,1,power_law_1.01,7.880684661865234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,1,power_law_1.01,10.69324188232422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,1,power_law_1.01,13.329177856445312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,1,power_law_1.01,16.2501953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,1,power_law_1.01,25.51404724121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,1,balanced,2.938938776652018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,1,power_law_1.01,50.96330871582031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,1,balanced,5.306597391764323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,1,balanced,9.361103693644205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,1,power_law_1.01,0.15219199657440186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,1,power_law_1.01,0.22168319225311278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,1,power_law_1.01,0.3048896074295044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,1,power_law_1.01,0.4261119842529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,1,power_law_1.01,0.6140543937683105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,1,power_law_1.01,0.985747241973877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,1,power_law_1.01,1.0763775825500488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,1,power_law_1.01,1.1039423942565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,1,power_law_1.01,1.1448320388793944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,1,power_law_1.01,1.1215680122375489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,1,power_law_1.01,1.197113609313965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,1,power_law_1.01,1.178444766998291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,1,power_law_1.01,1.2238271713256836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,1,power_law_1.01,0.25471360683441163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,1,power_law_1.01,1.2521471977233887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,1,power_law_1.01,0.44967041015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,1,power_law_1.01,1.2661567687988282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,1,power_law_1.01,0.5893631935119629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,1,power_law_1.01,1.2529279708862304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,1,power_law_1.01,0.9802495956420898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,1,power_law_1.01,1.3079360008239747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,1,power_law_1.01,1.0674367904663087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,1,power_law_1.01,1.3966400146484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,1,power_law_1.01,1.1393792152404785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.01,1.521331214904785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,1,power_law_1.01,1.1531583786010742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,1,power_law_1.01,0.05071359872817993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,1,power_law_1.01,1.1576576232910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.01,1.7359552383422852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,1,power_law_1.01,0.07293440103530884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,1,power_law_1.01,1.1952447891235352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.01,1.925619125366211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,1,power_law_1.01,0.0836031973361969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,1,power_law_1.01,1.2071231842041015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.01,2.3572671890258787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,1,power_law_1.01,0.15036159753799438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,1,power_law_1.01,1.2099904060363769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.01,2.7991552352905273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,1,power_law_1.01,0.2300031900405884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,1,power_law_1.01,1.249465560913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.01,3.657574462890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,1,power_law_1.01,0.38480000495910643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,1,power_law_1.01,1.247430419921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.01,4.444377517700195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,1,power_law_1.01,0.4422976016998291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,1,power_law_1.01,1.280998420715332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,1,power_law_1.01,0.4306431770324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,1,power_law_1.01,1.322982406616211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.01,6.140473556518555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,1,power_law_1.01,0.47466239929199217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,1,power_law_1.01,1.4377920150756835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.01,7.6801597595214846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,1,power_law_1.01,0.4762944221496582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,1,power_law_1.01,1.5464832305908203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,1,power_law_1.01,0.5292799949645997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,1,power_law_1.01,1.760812759399414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.01,9.303231811523437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,1,power_law_1.01,0.5682112216949463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,1,power_law_1.01,1.9824127197265624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,1,power_law_1.01,0.5164671897888183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,1,power_law_1.01,14.119308471679688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,1,power_law_1.01,2.396633529663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,1,power_law_1.01,0.5662208080291748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,1,power_law_1.01,2.7607744216918944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,1,power_law_1.01,0.5784895896911622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,1,power_law_1.01,3.6030654907226562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,1,power_law_1.01,27.127731323242188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,1,power_law_1.01,0.564524793624878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,1,power_law_1.01,4.436313629150391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,1,power_law_1.01,0.601420783996582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,1,power_law_1.01,6.049580764770508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,1,power_law_1.01,0.6330880165100098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.01,0.6532544136047364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,1,power_law_1.01,7.584140777587891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.01,1.5107456207275392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,1,power_law_1.01,9.27786865234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.01,1.7004287719726563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.01,1.125011157989502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,1,power_law_1.01,14.22558135986328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.01,1.2009535789489747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.01,1.35862398147583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,1,power_law_1.01,27.959527587890626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.01,1.6627328872680665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.01,2.249612808227539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.01,2.7434688568115235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.01,3.2998016357421873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,1,power_law_1.01,5.045580673217773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,1,power_law_1.01,9.505977630615234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,1,power_law_1.2,0.21973121166229248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,1,power_law_1.2,0.2420351982116699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,1,power_law_1.2,0.4556608200073242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,1,power_law_1.2,0.6709311962127685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,1,power_law_1.2,1.008236789703369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,1,power_law_1.2,1.8268415451049804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,1,power_law_1.2,2.0207040786743162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,1,power_law_1.2,2.196780776977539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,1,power_law_1.2,2.1992895126342775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,1,power_law_1.2,2.274553680419922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,1,power_law_1.2,2.366355133056641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,1,power_law_1.2,2.3646591186523436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,1,power_law_1.2,2.3733184814453123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,1,power_law_1.2,2.3815616607666015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,1,power_law_1.2,2.4592639923095705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,1,power_law_1.2,2.461881637573242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,1,power_law_1.2,2.6058687210083007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,1,power_law_1.2,2.8040960311889647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,1,power_law_1.2,2.976153564453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,1,power_law_1.2,3.357804870605469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,1,power_law_1.2,3.772243118286133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,1,power_law_1.2,4.436320114135742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,1,power_law_1.2,5.151699066162109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,1,power_law_1.2,6.512044525146484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,1,power_law_1.2,7.907186889648438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,1,power_law_1.2,10.71506576538086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,1,power_law_1.2,13.423295593261718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,1,power_law_1.2,16.434661865234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,1,power_law_1.2,24.71435546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,1,power_law_1.2,49.609310913085935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,1,power_law_1.2,0.1409343957901001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,1,power_law_1.2,0.159552001953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,1,power_law_1.2,0.26131839752197267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,1,power_law_1.2,0.3683648109436035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,1,power_law_1.2,0.5044288158416748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,1,power_law_1.2,1.0146047592163085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,1,power_law_1.2,1.0911552429199218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,1,power_law_1.2,1.140505599975586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,1,power_law_1.2,1.180947208404541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,1,power_law_1.2,0.15166079998016357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,1,power_law_1.2,1.2241663932800293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,1,power_law_1.2,0.17548160552978515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,1,power_law_1.2,0.27651200294494627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,1,power_law_1.2,1.2353856086730957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,1,power_law_1.2,0.4204991817474365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,1,power_law_1.2,1.2769408226013184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,1,power_law_1.2,0.5403135776519775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,1,power_law_1.2,1.30316162109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,1,power_law_1.2,0.9683327674865723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,1,power_law_1.2,1.283948802947998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,1,power_law_1.2,1.0081727981567383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,1,power_law_1.2,1.3254976272583008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,1,power_law_1.2,1.097702407836914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,1,power_law_1.2,1.3317888259887696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,1,power_law_1.2,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,1,power_law_1.2,1.1573247909545898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,1,power_law_1.2,1.3883328437805176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,1,power_law_1.2,0.053414398431777955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,1,power_law_1.2,1.1545344352722169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,1,power_law_1.2,1.5322688102722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,1,power_law_1.2,0.08303359746932984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,1,power_law_1.2,1.1929471969604493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,1,power_law_1.2,1.6068159103393556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,1,power_law_1.2,0.14821120500564575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,1,power_law_1.2,1.2109439849853516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,1,power_law_1.2,1.8353023529052734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,1,power_law_1.2,0.20810239315032958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,1,power_law_1.2,1.2113344192504882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,1,power_law_1.2,2.032588768005371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,1,power_law_1.2,0.3703743934631348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,1,power_law_1.2,1.2459839820861816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,1,power_law_1.2,2.46549129486084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,1,power_law_1.2,0.41987199783325196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,1,power_law_1.2,1.2571200370788573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,1,power_law_1.2,2.877804756164551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,1,power_law_1.2,0.4490047931671143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,1,power_law_1.2,1.2642815589904786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,1,power_law_1.2,3.733766555786133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,1,power_law_1.2,0.49271678924560547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,1,power_law_1.2,1.3310400009155274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,1,power_law_1.2,4.494700622558594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,1,power_law_1.2,0.5112383842468262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,1,power_law_1.2,1.432032012939453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,1,power_law_1.2,0.5413951873779297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.2,1.5529791831970214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,1,power_law_1.2,6.148108673095703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,1,power_law_1.2,0.5609216213226318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.2,1.7516799926757813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,1,power_law_1.2,7.710259246826172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,1,power_law_1.2,0.5418047904968262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.2,1.946169662475586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,1,power_law_1.2,0.5698239803314209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,1,power_law_1.2,9.319929504394532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.2,2.365951919555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,1,power_law_1.2,0.6130303859710693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.2,2.789753532409668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,1,power_law_1.2,14.238284301757812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,1,power_law_1.2,0.57292160987854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.2,3.554246520996094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,1,power_law_1.2,0.6034624099731445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.2,4.354406356811523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,1,power_law_1.2,0.649625587463379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,1,power_law_1.2,28.16866455078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.2,5.955161666870117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.2,0.6816703796386718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.2,1.5484864234924316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.2,7.515277099609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.2,1.8228160858154296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.2,9.092108917236327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.2,1.1475584030151367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.2,1.2215231895446776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,1,power_law_1.2,13.873989868164063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.2,1.4082112312316895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.2,1.7371967315673829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,1,power_law_1.2,26.471450805664062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.2,2.305388832092285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.2,2.812870407104492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.2,3.409708786010742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,1,power_law_1.2,5.105203247070312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,1,power_law_1.2,9.51720962524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,1,balanced,0.19634666045506796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,1,balanced,0.35283199946085614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,1,balanced,0.5747573375701904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,1,balanced,1.0136533578236897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,1,balanced,1.8485493659973145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,1,balanced,3.5032052993774414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,1,balanced,5.129205385843913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,1,balanced,5.183685302734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,1,balanced,5.12279478708903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,1,balanced,5.221690813700358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,1,balanced,5.247477213541667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,1,balanced,5.316752115885417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,1,balanced,5.313440004984538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,1,balanced,5.280384063720703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,1,balanced,5.356394449869792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,1,balanced,5.154597282409668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,1,balanced,5.348447799682617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,1,balanced,5.295674641927083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,1,balanced,5.391077041625977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,1,balanced,5.4032853444417315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,1,balanced,5.450368245442708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,1,balanced,5.605797449747722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,1,balanced,5.699861526489258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,1,balanced,6.048709233601888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,1,balanced,10.08738644917806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,1,balanced,9.220085144042969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,1,balanced,0.14336533347765604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,1,balanced,0.2077173391977946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,1,balanced,0.17553067207336426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,1,balanced,0.35035733381907147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,1,balanced,0.24728532632191977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,1,balanced,0.5786453485488892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,1,balanced,0.38415467739105225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,1,balanced,1.0229706764221191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,1,balanced,0.6435946623484293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,1,balanced,1.8592640558878581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,1,balanced,1.0143626530965169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,1,balanced,2.6944586435953775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,1,balanced,1.8416479428609211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,1,balanced,2.6571787198384604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,1,balanced,2.5848053296407065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,1,balanced,2.5828372637430825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,1,balanced,2.669642766316732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,1,balanced,2.5069066683451333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,1,balanced,2.6544853846232095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,1,balanced,2.470085302988688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,1,balanced,2.678442637125651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,1,balanced,2.4388319651285806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,1,balanced,2.6886345545450845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,1,balanced,14.017557779947916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,1,balanced,2.402341365814209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,1,balanced,2.7302347819010415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,1,power_law_1.01,0.2091264009475708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,1,balanced,2.3936586380004883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,1,balanced,2.7054611841837564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,1,power_law_1.01,0.347763204574585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,1,balanced,2.4496426582336426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,1,balanced,2.8130772908528647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,1,balanced,2.560266653696696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,1,power_law_1.01,0.5794432163238525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,1,balanced,2.7878828048706055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,1,balanced,2.5203679402669272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,1,power_law_1.01,0.7240575790405274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,1,balanced,2.7836427688598633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,1,balanced,2.540367921193441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,1,power_law_1.01,1.4156864166259766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,1,balanced,2.7810560862223306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,1,power_law_1.01,2.020102310180664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,1,balanced,2.570144017537435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,1,balanced,2.8755839665730796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,1,power_law_1.01,2.3813312530517576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,1,balanced,2.5833226839701333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,1,power_law_1.01,3.287993621826172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,1,balanced,2.8895626068115234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,1,balanced,2.5759414037068686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,1,power_law_1.01,3.39153938293457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,1,power_law_1.01,3.582297515869141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,1,balanced,2.733674685160319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,1,balanced,2.978917439778646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,1,power_law_1.01,3.6317249298095704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,1,balanced,19.78101348876953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,1,power_law_1.01,3.7426944732666017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,1,balanced,2.8046881357828775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,1,balanced,3.108229319254557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,1,power_law_1.01,4.034086227416992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,1,power_law_1.01,4.011161422729492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,1,power_law_1.01,4.222822570800782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,1,balanced,2.9144159952799478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,1,balanced,3.292463938395182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,1,power_law_1.01,4.2272895812988285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,1,power_law_1.01,4.319827270507813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,1,power_law_1.01,4.371603012084961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,1,balanced,3.277850786844889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,1,power_law_1.01,4.623686218261719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,1,balanced,3.6351839701334634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,1,power_law_1.01,5.121855926513672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,1,power_law_1.01,5.502803039550781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,1,power_law_1.01,6.356383895874023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,1,balanced,5.560085296630859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,1,balanced,5.637514750162761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,1,power_law_1.01,7.101593780517578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,1,power_law_1.01,8.515007781982423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,1,power_law_1.01,10.094745635986328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,1,power_law_1.01,13.01226806640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,1,balanced,5.778922398885091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,1,power_law_1.01,15.937254333496094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,1,balanced,28.205841064453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,1,balanced,5.715578715006511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,1,power_law_1.01,19.002149963378905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,1,power_law_1.01,27.639987182617187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,1,balanced,7.948895772298177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,1,power_law_1.01,50.586181640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,1,balanced,8.19755744934082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,1,balanced,10.859578450520834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,1,balanced,11.36832046508789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,1,balanced,15.987759908040365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,1,balanced,47.98906453450521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,1,balanced,16.939732869466145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,1,balanced,28.552459716796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,1,balanced,29.958948771158855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,1,power_law_1.01,0.13018239736557008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,1,power_law_1.01,0.20056960582733155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,1,power_law_1.01,0.32660479545593263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,1,power_law_1.01,0.40659198760986326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,1,power_law_1.01,0.7700479984283447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,1,power_law_1.01,1.1243583679199218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,1,power_law_1.01,1.3403776168823243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,1,power_law_1.01,1.7232704162597656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,1,power_law_1.01,1.7804031372070312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,1,power_law_1.01,1.856166458129883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,1,power_law_1.01,1.9306495666503907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,1,power_law_1.01,2.0560895919799806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,1,power_law_1.01,2.080998420715332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,1,power_law_1.01,2.137664031982422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,1,power_law_1.01,2.1915456771850588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,1,power_law_1.01,2.224473571777344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,1,power_law_1.01,2.287615966796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,1,power_law_1.01,2.4302207946777346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,1,power_law_1.01,2.5298879623413084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,1,power_law_1.01,2.776371192932129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,1,power_law_1.01,3.0096832275390626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,1,power_law_1.01,3.4724353790283202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,1,power_law_1.01,3.8803966522216795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,1,power_law_1.01,4.783404922485351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,1,power_law_1.01,0.18388479948043823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,1,power_law_1.01,5.762988662719726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,1,power_law_1.01,0.2447808027267456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,1,power_law_1.01,7.446080017089844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,1,power_law_1.01,0.401855993270874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,1,power_law_1.01,9.213203430175781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,1,power_law_1.01,0.45513601303100587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,1,power_law_1.01,0.8259455680847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,1,power_law_1.01,10.876294708251953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,1,power_law_1.01,1.1182527542114258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,1,power_law_1.01,1.3546560287475586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,1,power_law_1.01,16.208741760253908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,1,power_law_1.01,1.6610111236572265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,1,power_law_1.01,1.7848320007324219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,1,power_law_1.01,30.25287780761719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,1,power_law_1.01,1.785113525390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,1,power_law_1.01,1.9055551528930663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,1,power_law_1.01,1.9657279968261718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,1,power_law_1.01,2.0058496475219725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,1,power_law_1.01,2.0329280853271485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,1,power_law_1.01,2.090675163269043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,1,power_law_1.01,2.159219169616699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,1,power_law_1.01,2.243935966491699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,1,power_law_1.01,2.3261823654174805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.01,2.35032958984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,1,power_law_1.2,0.19144320487976074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,1,power_law_1.2,0.3344063997268677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.01,2.6015615463256836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,1,power_law_1.2,0.41621761322021483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.01,2.8524927139282226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,1,power_law_1.2,0.7377535820007324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.01,3.294220733642578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,1,power_law_1.2,1.243391990661621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.01,3.7274303436279297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,1,power_law_1.2,1.7817663192749023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.01,4.5714366912841795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,1,power_law_1.2,2.090675163269043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,1,power_law_1.01,5.463692855834961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,1,power_law_1.2,3.0052223205566406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,1,power_law_1.01,7.249874877929687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,1,power_law_1.2,3.2181888580322267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,1,power_law_1.2,3.3397377014160154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,1,power_law_1.01,8.905510711669923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,1,power_law_1.2,3.5839614868164062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,1,power_law_1.01,10.656281280517579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,1,power_law_1.2,3.7084800720214846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,1,power_law_1.2,0.13093760013580322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,1,power_law_1.01,15.710079956054688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,1,power_law_1.2,3.8102142333984377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,1,power_law_1.2,0.20262401103973388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,1,power_law_1.2,3.8996864318847657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,1,power_law_1.2,0.25004799365997316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,1,power_law_1.2,4.08026237487793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,1,power_law_1.01,29.232748413085936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,1,power_law_1.2,0.4182271957397461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,1,power_law_1.2,4.1881153106689455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,1,power_law_1.2,0.6840640068054199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,1,power_law_1.2,4.477497482299805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,1,power_law_1.2,0.9774144172668457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,1,power_law_1.2,4.643737411499023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,1,power_law_1.2,1.1778240203857422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,1,power_law_1.2,4.914022445678711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,1,power_law_1.2,1.6683135986328126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,1,power_law_1.2,5.451174545288086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,1,power_law_1.2,1.7285568237304687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,1,power_law_1.2,5.869401550292968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,1,power_law_1.2,1.819615936279297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,1,power_law_1.2,6.756671905517578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,1,power_law_1.2,1.8954048156738281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,1,power_law_1.2,7.417350769042969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,1,power_law_1.2,1.9378944396972657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,1,power_law_1.2,8.894316864013671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,1,power_law_1.2,2.0411840438842774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,1,power_law_1.2,2.163321685791016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,1,power_law_1.2,10.467001342773438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,1,power_law_1.2,2.2431615829467773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,1,power_law_1.2,13.263475036621093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,1,power_law_1.2,2.2634368896484376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,1,power_law_1.2,2.3296768188476564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,1,power_law_1.2,16.041510009765624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,1,power_law_1.2,2.521017646789551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,1,power_law_1.2,2.609516716003418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,1,power_law_1.2,18.863775634765624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,1,power_law_1.2,2.906892776489258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,1,power_law_1.2,3.1837568283081055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,1,power_law_1.2,27.740447998046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,1,power_law_1.2,3.7087871551513674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,1,power_law_1.2,4.156288146972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,1,power_law_1.2,50.359884643554686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,1,power_law_1.2,5.080204772949219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,1,power_law_1.2,5.9339134216308596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,1,power_law_1.2,7.680063629150391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,1,power_law_1.2,9.393138885498047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,1,power_law_1.2,11.051423645019531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,1,power_law_1.2,16.40709686279297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,1,power_law_1.2,31.143167114257814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,1,power_law_1.2,0.17403520345687867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,1,power_law_1.2,0.2350208044052124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,1,power_law_1.2,0.27392001152038575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,2,balanced,0.16060800353686014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,1,power_law_1.2,0.4779712200164795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,2,balanced,0.2399359941482544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,2,balanced,0.3956480026245117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,1,power_law_1.2,0.7218239784240723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,2,balanced,0.3671146631240845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,2,balanced,0.33316266536712646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,1,power_law_1.2,0.9600576400756836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,2,balanced,0.3493386507034302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,2,balanced,0.3408213456471761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,1,power_law_1.2,1.1875967979431152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,2,balanced,0.32872533798217773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,1,power_law_1.2,1.628486442565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,2,balanced,0.33113600810368854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,2,balanced,0.3665653467178345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,1,power_law_1.2,1.7063360214233398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,2,balanced,0.3358986775080363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,2,balanced,0.3323199947675069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,1,power_law_1.2,1.7807872772216797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,2,balanced,0.34091734886169434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,1,power_law_1.2,1.843257522583008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,2,balanced,0.2934826612472534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,2,balanced,0.2794133424758911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,1,power_law_1.2,1.9106496810913085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,2,balanced,0.27183467149734497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,2,balanced,0.2742026646931966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,1,power_law_1.2,2.015148735046387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,2,balanced,0.5011093219121298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,1,power_law_1.2,2.047417640686035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,2,balanced,0.4690026839574178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,2,balanced,0.6630186637242635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,1,power_law_1.2,2.090540885925293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,2,balanced,0.8617066542307535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,1,power_law_1.2,2.152895927429199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,2,balanced,1.2362453142801921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,1,power_law_1.2,2.199078369140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,1,power_law_1.2,2.4081472396850585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,2,balanced,1.5887680053710938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.2,2.523910331726074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,2,balanced,2.280282656351725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.2,2.8597312927246095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.2,3.0049407958984373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,2,balanced,2.9010238647460938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.2,3.569919967651367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.2,3.983001708984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,2,balanced,4.0793759028116865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.2,4.911801528930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,1,power_law_1.2,5.725785446166992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,2,balanced,0.10752532879511516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,2,balanced,5.224560101826985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,2,balanced,0.13938666383425394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,1,power_law_1.2,7.423545837402344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,2,balanced,0.19940267006556192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,2,balanced,0.1883253256479899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,1,power_law_1.2,9.114189147949219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,2,balanced,0.18968532482783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,2,balanced,0.18340800205866495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,2,balanced,0.21812800566355386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,2,balanced,6.374218622843425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,1,power_law_1.2,10.841241455078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,2,balanced,0.19340266784032187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,2,balanced,0.22038400173187256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,2,balanced,0.09133332967758179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,2,balanced,0.19841599464416504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,1,power_law_1.2,15.864256286621094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,2,balanced,0.12981333335240683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,2,balanced,0.20124799013137817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,2,balanced,0.19782400131225586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,2,balanced,0.2076266606648763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,2,balanced,0.19726399580637613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,2,balanced,0.18632533152898154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,2,balanced,0.20510399341583252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,2,balanced,9.698176066080729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,1,power_law_1.2,29.203155517578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,2,balanced,0.18220800161361694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,2,balanced,0.19221866130828857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,2,balanced,0.1797973314921061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,2,balanced,0.2037173310915629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,2,balanced,0.17519466082255045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,2,balanced,0.20121065775553384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,2,balanced,0.17918932437896729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,2,balanced,0.20247999827067056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,2,balanced,0.1999573310216268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,2,balanced,0.31063467264175415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,2,balanced,0.18773333231608072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,2,balanced,0.26788800954818726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,2,balanced,0.2081973354021708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,2,balanced,0.3657919963200887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,2,balanced,0.17733865976333618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,2,balanced,0.175818661848704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,2,balanced,0.45740799109141034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,2,balanced,0.15923733512560526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,2,balanced,0.6388853391011556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,2,balanced,0.15382933616638184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,2,balanced,18.533626556396484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,2,balanced,0.1414346694946289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,2,balanced,0.8034026622772217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,2,balanced,0.2782026727994283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,2,balanced,1.143957297007243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,2,balanced,0.21635733048121134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,2,balanced,0.28390934069951373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,2,balanced,0.34942400455474854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,2,balanced,1.481760025024414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,2,balanced,0.49750932057698566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,2,balanced,0.6421013275782267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,2,balanced,2.142218589782715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,2,balanced,0.9270239671071371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,2,balanced,1.2041707038879395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,2,balanced,2.794992129007975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,2,balanced,1.7804320653279622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,2,balanced,3.4464852015177407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,2,balanced,2.3645386695861816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,2,balanced,5.383130391438802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,2,balanced,2.9400691986083984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,2,balanced,4.659765243530273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,2,balanced,10.611370722452799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,2,balanced,9.171503702799479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,2,balanced,0.033861334125200905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,2,balanced,0.03468266626199087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,2,balanced,0.032960000137488045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,2,balanced,0.039450667798519135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,2,balanced,0.04821333289146423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,2,balanced,0.04855466882387797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,2,power_law_1.01,0.19870079755783082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,2,balanced,0.04684266448020935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,2,power_law_1.01,0.20410239696502686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,2,balanced,0.050426666935284935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,2,power_law_1.01,0.23230719566345215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,2,balanced,0.06618666648864746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,2,power_law_1.01,0.3645888090133667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,2,balanced,0.053599998354911804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,2,power_law_1.01,0.3576191902160645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,2,balanced,0.058117335041364036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,2,power_law_1.01,0.3537600040435791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,2,balanced,0.11713600158691406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,2,power_law_1.01,0.3367167949676514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,2,balanced,0.11720533172289531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,2,power_law_1.01,0.3456255912780762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,2,balanced,0.13640000422795615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,2,power_law_1.01,0.33240320682525637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,2,balanced,0.11591999729474385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,2,power_law_1.01,0.3353856086730957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,2,power_law_1.01,0.3257152080535889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,2,balanced,0.08416000008583069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,2,power_law_1.01,0.36246399879455565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,2,balanced,0.09569600224494934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,2,power_law_1.01,0.3861248016357422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,2,balanced,0.13209600249926248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,2,power_law_1.01,0.3869119882583618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,2,balanced,0.15686399737993875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,2,power_law_1.01,0.4357439994812012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,2,balanced,0.22126932938893637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,2,power_law_1.01,0.4879936218261719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,2,power_law_1.01,0.5234623908996582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,2,balanced,0.280074675877889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,2,power_law_1.01,0.635539197921753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,2,power_law_1.01,0.8156800270080566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,2,balanced,0.39708268642425537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,2,power_law_1.01,0.9434047698974609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,2,power_law_1.01,1.1838015556335448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,2,balanced,0.5013173421223959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,2,power_law_1.01,1.9446975708007812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,2,power_law_1.01,2.3434368133544923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,2,balanced,0.7435093720753988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,2,power_law_1.01,3.4110782623291014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,2,power_law_1.01,4.770854568481445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,2,power_law_1.01,5.516479873657227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,2,balanced,0.9631199836730957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,2,power_law_1.01,8.380166625976562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,2,power_law_1.01,8.71364517211914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,2,balanced,1.4220587412516277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,2,power_law_1.01,12.770861053466797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,2,power_law_1.01,22.538706970214843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,2,balanced,1.8643520673116047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,2,power_law_1.01,0.11247999668121338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,2,balanced,2.325904051462809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,2,power_law_1.01,0.12184319496154786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,2,power_law_1.01,0.15646719932556152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,2,power_law_1.01,0.20378880500793456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,2,power_law_1.01,0.2116096019744873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,2,power_law_1.01,0.20713601112365723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,2,power_law_1.01,0.2019263982772827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,2,power_law_1.01,0.20200319290161134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,2,power_law_1.01,0.2035968065261841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,2,power_law_1.01,0.10054399967193603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,2,power_law_1.01,0.19601919651031494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,2,balanced,3.661381403605143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,2,power_law_1.01,0.12878719568252564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,2,power_law_1.01,0.1931712031364441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,2,power_law_1.01,0.15233919620513917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,2,power_law_1.01,0.2134399890899658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,2,power_law_1.01,0.19676159620285033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,2,power_law_1.01,0.22755839824676513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,2,power_law_1.01,0.19685759544372558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,2,power_law_1.01,0.22872319221496581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,2,power_law_1.01,0.1999295949935913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,2,power_law_1.01,0.2583807945251465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,2,power_law_1.01,0.18562560081481932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,2,power_law_1.01,0.27886080741882324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,2,power_law_1.01,0.1985152006149292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,2,power_law_1.01,0.30117120742797854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,2,power_law_1.01,0.196563196182251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,2,power_law_1.01,0.3207808017730713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,2,power_law_1.01,0.18341120481491088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,2,power_law_1.01,0.44537601470947263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,2,power_law_1.01,0.19120639562606812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,2,power_law_1.01,0.5780735969543457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,2,power_law_1.01,0.178438401222229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,2,power_law_1.01,0.6481535911560059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,2,power_law_1.01,0.19356800317764283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,2,power_law_1.01,0.9870847702026367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,2,power_law_1.01,0.20545918941497804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,2,power_law_1.01,1.0930879592895508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,2,power_law_1.01,0.23705599308013917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,2,power_law_1.01,1.6278335571289062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,2,power_law_1.01,0.2500799894332886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,2,power_law_1.01,2.4392127990722656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,2,balanced,7.282005310058594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,2,power_law_1.01,0.25498878955841064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,2,power_law_1.01,3.3556671142578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,2,power_law_1.01,0.3053375959396362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.01,0.37458560466766355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,2,power_law_1.01,4.328531265258789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.01,0.48583040237426756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,2,power_law_1.01,4.934431838989258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.01,0.5778175830841065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,2,power_law_1.01,7.918112182617188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.01,0.693120002746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.01,1.1002880096435548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,2,power_law_1.01,18.02857666015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.01,1.1965567588806152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.01,1.8959615707397461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.01,2.5463680267333983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.01,3.5297985076904297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,2,power_law_1.01,4.037126541137695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,2,power_law_1.01,7.713581085205078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,2,power_law_1.01,13.018098449707031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,2,power_law_1.01,0.03180800080299377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,2,power_law_1.01,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,2,power_law_1.01,0.03240959942340851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,2,power_law_1.01,0.03772799968719483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,2,power_law_1.01,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,2,power_law_1.01,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,2,power_law_1.01,0.04446719884872437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,2,power_law_1.01,0.048556798696517946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,2,power_law_1.01,0.06289280056953431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,2,power_law_1.01,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,2,power_law_1.01,0.05733759999275208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,2,power_law_1.01,0.1153216004371643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,2,power_law_1.01,0.11557120084762573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,2,power_law_1.01,0.13248640298843384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,2,power_law_1.01,0.11178879737854004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,2,power_law_1.01,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,2,power_law_1.01,0.09192960262298584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,2,power_law_1.01,0.12980480194091798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.01,0.15459200143814086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.01,0.2193471908569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.01,0.2780224084854126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.01,0.39495680332183836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.01,0.5018943786621094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.01,0.7340095996856689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.01,0.9623744010925293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.01,1.4158592224121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.01,1.8630207061767579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,2,power_law_1.01,2.315091133117676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,2,power_law_1.2,0.20602240562438964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,2,power_law_1.01,3.681932830810547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,2,power_law_1.2,0.24712960720062255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,2,power_law_1.01,7.296537780761719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,2,power_law_1.2,0.2531071901321411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,2,power_law_1.2,0.34749441146850585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,2,power_law_1.2,0.3469887971878052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,2,power_law_1.2,0.3419903993606567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,2,power_law_1.2,0.33912320137023927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,2,power_law_1.2,0.34339840412139894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,2,power_law_1.2,0.12309119701385499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,2,power_law_1.2,0.3517184019088745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,2,power_law_1.2,0.14384000301361083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,2,power_law_1.2,0.3318720102310181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,2,power_law_1.2,0.12071039676666259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,2,power_law_1.2,0.32831358909606934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,2,power_law_1.2,0.19181439876556397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,2,power_law_1.2,0.3579200029373169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,2,power_law_1.2,0.19360640048980712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,2,power_law_1.2,0.38600959777832033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,2,power_law_1.2,0.4243584156036377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,2,power_law_1.2,0.4442304134368896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,2,power_law_1.2,0.46727681159973145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,2,power_law_1.2,0.1362239956855774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,2,power_law_1.2,0.5421247959136963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,2,power_law_1.2,0.11964160203933716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,2,power_law_1.2,0.666758394241333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,2,power_law_1.2,0.1572160005569458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,2,power_law_1.2,0.8598655700683594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,2,power_law_1.2,0.20523519515991212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,2,power_law_1.2,1.0856575965881348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,2,power_law_1.2,0.20572800636291505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,2,power_law_1.2,1.4350208282470702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,2,power_law_1.2,0.19799679517745972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,2,power_law_1.2,1.9343040466308594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,2,power_law_1.2,0.19646719694137574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,2,power_law_1.2,2.615564727783203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,2,power_law_1.2,0.1941248059272766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,2,power_law_1.2,3.550105667114258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,2,power_law_1.2,0.19211519956588746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,2,power_law_1.2,0.1925055980682373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,2,power_law_1.2,4.705887985229492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,2,power_law_1.2,0.19159679412841796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,2,power_law_1.2,5.859648132324219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,2,power_law_1.2,0.20492160320281982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,2,power_law_1.2,9.127986907958984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,2,power_law_1.2,0.2218559980392456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,2,power_law_1.2,0.19130239486694336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,2,power_law_1.2,11.348121643066406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,2,power_law_1.2,0.23002240657806397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,2,power_law_1.2,0.24344959259033203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,2,power_law_1.2,13.371250915527344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,2,power_law_1.2,0.2668735980987549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,2,power_law_1.2,0.3247744083404541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.2,0.3854527950286865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.2,0.4887040138244629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,2,power_law_1.2,29.538540649414063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.2,0.5179967880249023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.2,0.8539135932922364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.2,1.0426048278808593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.2,1.546342372894287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.2,1.8691520690917969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.2,2.983647918701172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.2,3.3836032867431642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,2,power_law_1.2,4.807020950317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,2,power_law_1.2,7.773446655273437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,2,power_law_1.2,14.503187561035157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,2,power_law_1.2,0.03118079900741577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,2,power_law_1.2,0.03121280074119568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,2,power_law_1.2,0.030969598889350893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,2,power_law_1.2,0.03702400028705597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,2,power_law_1.2,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,2,power_law_1.2,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,2,power_law_1.2,0.04371840059757233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,2,power_law_1.2,0.04826880097389221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,2,power_law_1.2,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,2,power_law_1.2,0.05134080052375793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,2,power_law_1.2,0.05430399775505066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,2,power_law_1.2,0.11463040113449097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,2,power_law_1.2,0.11587200164794922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,2,power_law_1.2,0.13159680366516113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,2,power_law_1.2,0.10820480585098266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,2,power_law_1.2,0.08335360288619995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,2,power_law_1.2,0.09171839952468872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,2,power_law_1.2,0.12847360372543334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.2,0.15574400424957274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.2,0.21946239471435547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.2,0.2776128053665161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.2,0.39422080516815183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.2,0.5029952049255371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.2,0.731609582901001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.2,0.9620608329772949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.2,1.4104640007019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.2,1.8609216690063477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,2,power_law_1.2,2.3130367279052733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,2,power_law_1.2,3.663577651977539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,1,balanced,0.14030399918556213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,2,power_law_1.2,7.2982017517089846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,1,balanced,0.19449599583943686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,1,balanced,0.35340265432993573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,1,balanced,0.2996426622072856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,1,balanced,0.3385546604792277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,1,balanced,0.32817065715789795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,1,balanced,0.3173440098762512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,1,balanced,0.3291520078976949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,1,balanced,0.31754134098688763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,1,balanced,0.3344159921010335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,1,balanced,0.31989334026972455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,1,balanced,0.3043893376986186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,1,balanced,0.2933013240496318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,1,balanced,0.2881866693496704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,1,balanced,0.27612266937891644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,1,balanced,0.2765439947446187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,1,balanced,0.28278932968775433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,1,balanced,0.49183468023935956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,1,balanced,0.444597323735555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,1,balanced,0.0962666670481364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,1,balanced,0.12772799531618753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,1,balanced,0.6056319872538248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,1,balanced,0.21318932374318442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,1,balanced,0.7374933560689291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,1,balanced,0.1904639999071757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,1,balanced,0.17079466581344604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,1,balanced,1.0200106302897136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,1,balanced,0.18791999419530234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,1,balanced,0.21558932463328043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,1,balanced,0.200272003809611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,1,balanced,1.2939039866129558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,1,balanced,0.1966666579246521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,1,balanced,0.19914132356643677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,1,balanced,1.845685323079427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,1,balanced,0.19762667020161948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,1,balanced,0.18079467614491782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,2,power_law_1.2,0.1957695960998535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,1,balanced,0.17428267002105713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,2,power_law_1.2,0.20459520816802979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,1,balanced,0.17430400848388672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,1,balanced,2.386250654856364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,2,power_law_1.2,0.19641599655151368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,1,balanced,0.17722133795420328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,2,power_law_1.2,0.20714240074157714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,1,balanced,0.1769919991493225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,2,power_law_1.2,0.20123519897460937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,1,balanced,0.18323200941085815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,2,power_law_1.2,0.19468799829483033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,1,balanced,3.4937705993652344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,1,balanced,0.2775680025418599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,2,power_law_1.2,0.23399679660797118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,1,balanced,0.26021866003672284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,2,power_law_1.2,0.22161920070648194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,2,power_law_1.2,0.23153278827667237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,1,balanced,0.34354666868845624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,2,power_law_1.2,0.24637439250946044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,1,balanced,0.42186665534973145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,1,balanced,4.60372257232666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,2,power_law_1.2,0.2815743923187256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,2,power_law_1.2,0.29687039852142333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,1,balanced,0.5887680053710938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,2,power_law_1.2,0.32760961055755616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,2,power_law_1.2,0.4174528121948242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,1,balanced,0.7562133471171061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,2,power_law_1.2,0.5772416114807128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,1,balanced,5.68283208211263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,2,power_law_1.2,0.690118408203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,2,power_law_1.2,1.017523193359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,1,balanced,1.0921066602071126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,2,power_law_1.2,1.1859647750854492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,2,power_law_1.2,1.5272576332092285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,1,balanced,1.4311572710673015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,2,power_law_1.2,2.566783905029297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,2,power_law_1.2,3.383647918701172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,1,balanced,9.220698674519857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,2,power_law_1.2,4.424262237548828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,2,power_law_1.2,6.001696014404297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,1,balanced,2.106501261393229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,2,power_law_1.2,8.735404968261719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,2,power_law_1.2,16.310823059082033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,1,balanced,2.790719985961914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,1,balanced,18.473962148030598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,1,balanced,3.4552958806355796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,1,balanced,5.48361078898112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,1,balanced,0.08081066608428955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,1,balanced,0.12102400263150533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,1,balanced,0.1899413267771403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,1,balanced,0.2030186653137207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,1,balanced,0.19383466243743896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,1,balanced,0.17882666985193887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,1,balanced,0.20654932657877603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,1,balanced,0.19347200791041055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,1,balanced,0.1940106749534607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,1,balanced,0.200762669245402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,1,balanced,0.17730132738749185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,1,balanced,0.17571200927098593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,1,balanced,0.17613333463668823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,1,balanced,0.17094399531682333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,1,balanced,0.1658506691455841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,1,balanced,0.15121600031852722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,1,balanced,0.14965333541234335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,1,balanced,0.27428267399470013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,1,balanced,10.913028717041016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,1,balanced,0.22131733099619547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,1,balanced,0.2935626705487569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,1,balanced,0.36900798479715985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,1,balanced,0.5129546721776327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,1,balanced,0.662282665570577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,1,balanced,0.95414932568868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,1,balanced,1.2580426534016926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,1,balanced,1.8657066027323406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,1,balanced,2.443333307902018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,1,balanced,3.045248031616211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,1,balanced,4.84603214263916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,1,balanced,0.02362666775782903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,1,balanced,0.02518400053183238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,1,balanced,0.02605866640806198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,1,balanced,0.028853334486484528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,1,balanced,0.03233599911133448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,1,balanced,0.04307733476161957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,1,balanced,0.03880000114440918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,1,balanced,0.04310933252175649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,1,balanced,0.03862933317820231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,1,balanced,0.041402667760849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,1,balanced,0.04507733384768168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,1,power_law_1.01,0.1288831949234009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,1,balanced,9.765024185180664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,1,balanced,0.04454400142033895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,1,power_law_1.01,0.17223680019378662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,1,balanced,0.04830400149027506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,1,power_law_1.01,0.1879487991333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,1,balanced,0.05157866577307383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,1,power_law_1.01,0.3090496063232422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,1,balanced,0.09822932879130046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,1,power_law_1.01,0.34257280826568604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,1,balanced,0.11609066526095073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,1,power_law_1.01,0.33352320194244384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,1,balanced,0.13160000244776407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,1,power_law_1.01,0.33868160247802737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,1,power_law_1.01,0.33333759307861327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,1,balanced,0.08398399750391643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,1,power_law_1.01,0.31040639877319337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,1,balanced,0.09409067034721375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,1,power_law_1.01,0.332473611831665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,1,balanced,0.13060800234476724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,1,power_law_1.01,0.34057600498199464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,1,power_law_1.01,0.3350591897964478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,1,balanced,0.16402666767438254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,1,power_law_1.01,0.33546879291534426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,1,balanced,0.22818134228388467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,1,power_law_1.01,0.3538624048233032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,1,power_law_1.01,0.3978303909301758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,1,balanced,0.28774933020273846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,1,power_law_1.01,0.40198397636413574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,1,power_law_1.01,0.4320831775665283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,1,balanced,0.41309865315755206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,1,power_law_1.01,0.5112383842468262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,1,power_law_1.01,0.5911871910095214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,1,power_law_1.01,0.7232192039489747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,1,balanced,0.5389386812845866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,1,power_law_1.01,0.8799360275268555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,1,power_law_1.01,1.1912256240844727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,1,power_law_1.01,1.4721280097961427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,1,power_law_1.01,2.0412607192993164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,1,power_law_1.01,0.09002879858016968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,1,balanced,0.7837759653727213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,1,power_law_1.01,2.6013887405395506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,1,power_law_1.01,0.1115455985069275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,1,power_law_1.01,3.681350326538086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,1,power_law_1.01,0.09946879744529724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,1,power_law_1.01,0.19722880125045777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,1,power_law_1.01,4.800281524658203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,1,power_law_1.01,0.1787392020225525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,1,power_law_1.01,5.951980972290039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,1,balanced,1.031392018000285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,1,power_law_1.01,0.18734079599380493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,1,power_law_1.01,0.19768320322036742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,1,power_law_1.01,9.25696029663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,1,power_law_1.01,0.19685120582580568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,1,power_law_1.01,0.18376959562301637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,1,power_law_1.01,18.691973876953124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,1,power_law_1.01,0.18777600526809693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,1,power_law_1.01,0.19116159677505493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,1,balanced,1.2789440155029297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,1,power_law_1.01,0.194758403301239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,1,power_law_1.01,0.20253438949584962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,1,power_law_1.01,0.20195839405059815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,1,power_law_1.01,0.22263040542602539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,1,power_law_1.01,0.23955199718475342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,1,power_law_1.01,0.2560256004333496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,1,power_law_1.01,0.2960959911346436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,1,power_law_1.01,0.36046719551086426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,1,balanced,2.0175894101460776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,1,power_law_1.01,0.44851198196411135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,1,power_law_1.01,0.5074495792388916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,1,power_law_1.01,0.6754816055297852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,1,power_law_1.01,0.8537088394165039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,1,power_law_1.01,1.1979647636413575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,1,power_law_1.01,1.5178175926208497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,1,power_law_1.01,2.198534393310547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,1,power_law_1.01,2.9029567718505858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,1,power_law_1.01,3.5500415802001952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,1,power_law_1.01,5.571564865112305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,1,power_law_1.01,10.929945373535157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,1,balanced,3.979615847269694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,1,power_law_1.01,0.08840960264205933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,1,power_law_1.01,0.11255040168762206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,1,power_law_1.01,0.11687040328979492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,1,power_law_1.01,0.18773119449615477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,1,power_law_1.01,0.19367680549621583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,1,power_law_1.01,0.19370239973068237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,1,power_law_1.01,0.1896448016166687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,1,power_law_1.01,0.18938239812850952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,1,power_law_1.01,0.18569600582122803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,1,power_law_1.01,0.19468159675598146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,1,power_law_1.01,0.18665599822998047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,1,power_law_1.01,0.18992639780044557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,1,power_law_1.01,0.1918015956878662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,1,power_law_1.01,0.20378239154815675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,1,power_law_1.01,0.2286144018173218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,1,power_law_1.01,0.21807360649108887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,1,power_law_1.01,0.25742719173431394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,1,power_law_1.01,0.29459199905395506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.01,0.309497594833374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.01,0.40090241432189944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.01,0.45444478988647463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.01,0.6157951831817627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.01,0.7458047866821289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.01,1.05163516998291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.01,1.3451199531555176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.01,1.9322687149047852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.01,2.518764877319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.01,3.0968511581420897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,1,power_law_1.01,4.90332145690918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,1,power_law_1.01,9.800857543945312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,1,power_law_1.01,0.023398399353027344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,1,power_law_1.01,0.02399359941482544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,1,power_law_1.01,0.025241601467132568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,1,power_law_1.01,0.027852800488471986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,1,power_law_1.01,0.030969598889350893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,1,power_law_1.01,0.04289920032024384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,1,power_law_1.01,0.037599998712539676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,1,power_law_1.01,0.042656001448631284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,1,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,1,power_law_1.01,0.04132480025291443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,1,power_law_1.01,0.04321280121803284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,1,power_law_1.01,0.04268159866333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,1,power_law_1.01,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,1,power_law_1.01,0.04997760057449341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,1,power_law_1.01,0.0971455991268158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,1,power_law_1.01,0.11315200328826905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,1,power_law_1.01,0.1305343985557556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,1,power_law_1.01,0.08416000008583069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.01,0.09240319728851318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.01,0.12951040267944336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.01,0.1624511957168579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,1,power_law_1.2,0.08545920252799988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,1,power_law_1.2,0.14094079732894899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.01,0.22739200592041015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,1,power_law_1.2,0.09292160272598267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,1,power_law_1.2,0.16329599618911744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.01,0.28801279067993163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,1,power_law_1.2,0.12672640085220338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,1,power_law_1.2,0.19687039852142335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.01,0.41461758613586425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,1,power_law_1.2,0.18838399648666382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,1,power_law_1.2,0.3079744100570679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.01,0.5369088172912597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,1,power_law_1.2,0.3234560012817383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,1,power_law_1.2,0.19233920574188232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.01,0.7838399887084961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,1,power_law_1.2,0.33812479972839354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,1,power_law_1.2,0.1831104040145874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.01,1.0299136161804199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,1,power_law_1.2,0.32870399951934814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,1,power_law_1.2,0.19167360067367553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.01,1.2789759635925293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,1,power_law_1.2,0.32434558868408203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,1,power_law_1.2,0.1938688039779663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,1,power_law_1.01,2.016921615600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,1,power_law_1.2,0.3245055913925171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,1,power_law_1.2,0.18954240083694457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,1,power_law_1.2,0.3167871952056885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,1,power_law_1.01,3.9819583892822266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,1,power_law_1.2,0.18830080032348634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,1,power_law_1.2,0.3302272081375122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,1,power_law_1.2,0.18830080032348634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,1,power_law_1.2,0.3469376087188721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,1,power_law_1.2,0.20274560451507567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,1,power_law_1.2,0.3355711936950684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,1,power_law_1.2,0.204256010055542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,1,power_law_1.2,0.35731201171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,1,power_law_1.2,0.2107072114944458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,1,power_law_1.2,0.399072003364563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,1,power_law_1.2,0.22763519287109374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,1,power_law_1.2,0.41168642044067383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,1,power_law_1.2,0.2314687967300415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,1,power_law_1.2,0.40842242240905763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,1,power_law_1.2,0.2619839906692505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,1,power_law_1.2,0.5005119800567627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,1,power_law_1.2,0.29675519466400146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,1,power_law_1.2,0.6070911884307861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,1,power_law_1.2,0.33755519390106203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,1,power_law_1.2,0.7411007881164551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,1,power_law_1.2,0.4371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,1,power_law_1.2,0.8947903633117675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,1,power_law_1.2,0.49188480377197263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,1,power_law_1.2,1.1748160362243651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,1,power_law_1.2,0.6743167877197266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,1,power_law_1.2,1.4471039772033691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,1,power_law_1.2,0.8333632469177246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,1,power_law_1.2,2.023084831237793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,1,power_law_1.2,1.1651904106140136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,1,power_law_1.2,2.5932031631469727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,1,power_law_1.2,1.5101759910583497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,1,power_law_1.2,3.686489486694336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,1,power_law_1.2,2.1976703643798827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,1,power_law_1.2,4.7442176818847654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,1,power_law_1.2,2.8527807235717773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,1,power_law_1.2,5.880326461791992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,1,power_law_1.2,3.5258689880371095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,1,power_law_1.2,9.33526382446289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,1,power_law_1.2,5.5031169891357425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,1,power_law_1.2,18.21002197265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,1,power_law_1.2,10.995782470703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,1,power_law_1.2,0.08760960102081299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,1,power_law_1.2,0.09711359739303589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,1,power_law_1.2,0.11304320096969604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,1,power_law_1.2,0.19522559642791748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,1,power_law_1.2,0.1874176025390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,1,power_law_1.2,0.19057919979095458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,1,power_law_1.2,0.1922943949699402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,1,power_law_1.2,0.19489920139312744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,1,power_law_1.2,0.1834496021270752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,1,power_law_1.2,0.18751360177993776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,1,power_law_1.2,0.18366719484329225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,1,power_law_1.2,0.18882559537887572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,1,power_law_1.2,0.20540800094604492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,1,power_law_1.2,0.20556159019470216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,1,power_law_1.2,0.21779201030731202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,1,power_law_1.2,0.2182784080505371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,1,power_law_1.2,0.24360320568084717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,1,power_law_1.2,0.27118079662322997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.2,0.2971776008605957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.2,0.38718719482421876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.2,0.438431978225708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.2,0.5926464080810547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.2,0.7525055885314942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.2,1.0310784339904786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.2,1.335264015197754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.2,1.9230783462524415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,1,power_law_1.2,0.023686400055885314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.2,2.5214719772338867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,1,power_law_1.2,0.023423999547958374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.2,3.102828788757324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,1,power_law_1.2,0.024876800179481507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,1,power_law_1.2,0.026521599292755126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,1,power_law_1.2,4.919750213623047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,1,power_law_1.2,0.030483201146125793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,1,power_law_1.2,9.705363464355468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,1,power_law_1.2,0.04177919924259186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,1,power_law_1.2,0.0383103996515274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,1,power_law_1.2,0.04195840060710907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,1,power_law_1.2,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,1,power_law_1.2,0.04105600118637085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,1,power_law_1.2,0.04348160028457641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,1,power_law_1.2,0.042931199073791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,1,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,1,power_law_1.2,0.04987519979476929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,1,power_law_1.2,0.09787520170211791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,1,power_law_1.2,0.11320960521697998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,1,power_law_1.2,0.1312000036239624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,1,power_law_1.2,0.08188160061836243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.2,0.0919871985912323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.2,0.12899199724197388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,2,balanced,0.21530133485794067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.2,0.16302720308303834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,2,balanced,0.15101333459218344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,2,balanced,0.2952959934870402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.2,0.22819199562072753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,2,balanced,0.18037867546081543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,2,balanced,0.5246400038401285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.2,0.28771200180053713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,2,balanced,0.3204853336016337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,2,balanced,0.5686666568120321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,2,balanced,0.31536000967025757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.2,0.4137983798980713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,2,balanced,0.6014399925867716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,2,balanced,0.3547360102335612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,2,balanced,0.5826826492945353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.2,0.5377600193023682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,2,balanced,0.3116213281949361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,2,balanced,0.5930666526158651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.2,0.7839744091033936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,2,balanced,0.309279998143514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,2,balanced,0.5698346694310507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.2,1.0314623832702636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,2,balanced,0.32974932591120404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,2,balanced,0.5479199886322021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.2,1.2787263870239258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,2,balanced,0.32492266098658246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,2,balanced,0.5293226639429728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,1,power_law_1.2,2.01147518157959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,2,balanced,0.30371199051539105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,2,balanced,0.5241173505783081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,2,balanced,0.3242186705271403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,2,balanced,0.5141653219858805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,1,power_law_1.2,3.981689453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,2,balanced,0.30934399366378784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,2,balanced,0.5166666507720947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,2,balanced,0.3121013243993123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,2,balanced,0.5017600059509277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,2,balanced,0.2684640089670817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,2,balanced,0.4780160188674927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,2,balanced,0.2824479937553406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,2,balanced,0.46410131454467773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,2,balanced,0.2725920081138611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,2,balanced,0.4637920061747233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,2,balanced,0.2751306692759196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,2,balanced,0.8214080333709717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,2,balanced,0.49210135142008465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,2,balanced,0.7831306457519531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,2,balanced,0.4473866621653239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,2,balanced,1.0746453603108723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,2,balanced,0.6120373407999674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,2,balanced,1.4404266675313313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,2,balanced,0.7932853698730469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,2,balanced,2.08787202835083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,2,balanced,1.1151200135548909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,2,balanced,2.758053461710612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,2,balanced,1.4458613395690918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,2,balanced,4.086949348449707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,2,balanced,2.071845372517904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,2,balanced,5.31981881459554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,2,balanced,2.597226619720459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,2,balanced,7.734538396199544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,2,balanced,3.673930803934733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,2,balanced,10.115370432535807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,2,balanced,4.7357972462972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,2,balanced,12.171311696370443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,2,balanced,5.721200307210286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,2,balanced,8.751957575480143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,2,balanced,17.666709899902344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,2,balanced,0.13079466422398886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,2,balanced,0.18966933091481528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,2,balanced,0.3461120128631592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,2,balanced,0.32286399602890015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,2,balanced,0.32305065790812176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,2,balanced,0.3068693280220032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,2,balanced,0.30769066015879315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,2,balanced,0.31775466601053876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,2,balanced,0.31752532720565796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,2,balanced,0.3022666573524475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,2,balanced,16.917850494384766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,2,balanced,0.3132586677869161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,2,balanced,0.30447999636332196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,2,balanced,0.2742080092430115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,2,balanced,32.5921376546224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,2,balanced,0.25917865832646686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,2,balanced,0.24576000372568765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,2,balanced,0.223962664604187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,2,balanced,0.19714132944742838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,2,balanced,0.41837334632873535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,2,balanced,0.3258133331934611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,2,balanced,0.4412746826807658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,2,balanced,0.5611040194829305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,2,balanced,0.7903839747111002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,2,balanced,0.047930667797724404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,2,balanced,0.05085866649945577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,2,balanced,1.0256853103637695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,2,balanced,0.05119466781616211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,2,balanced,0.05941333373387655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,2,balanced,1.4848747253417969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,2,balanced,0.0900266667207082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,2,balanced,0.09081600109736125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,2,balanced,0.07358933488527934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,2,balanced,1.9332906405131023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,2,balanced,0.09479467074076335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,2,balanced,0.09070932865142822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,2,balanced,0.07998933394749959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,2,balanced,0.10547733306884766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,2,balanced,2.855034510294596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,2,balanced,0.1808799902598063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,2,balanced,0.1801919937133789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,2,balanced,0.2161066730817159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,2,balanced,0.14802666505177817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,2,balanced,3.7616958618164062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,2,balanced,0.1234933336575826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,2,balanced,0.1593546668688456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,2,balanced,0.20868800083796182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,2,balanced,0.26849599679311115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,2,balanced,4.655002593994141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,2,balanced,0.37620266278584796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,2,balanced,0.4950186808904012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,2,balanced,0.7068800131479899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,2,balanced,0.9301066398620605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,2,balanced,7.4099680582682295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,2,balanced,1.3667306900024414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,2,balanced,1.8082399368286133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,2,balanced,2.6786985397338867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,2,balanced,14.688735961914062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,2,balanced,3.5435733795166016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,2,balanced,4.428378740946452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,2,balanced,7.035258611043294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,2,power_law_1.01,0.15133440494537354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,2,power_law_1.01,0.2836287975311279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,2,power_law_1.01,0.19574400186538696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,2,power_law_1.01,0.35418241024017333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,2,power_law_1.01,0.25233280658721924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,2,power_law_1.01,0.4398399829864502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,2,power_law_1.01,0.31760640144348146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,2,power_law_1.01,0.5886335849761963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,2,power_law_1.01,0.31979520320892335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,2,power_law_1.01,0.5643392086029053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,2,power_law_1.01,0.3038464069366455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,2,power_law_1.01,0.5391679763793945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,2,power_law_1.01,0.30032639503479003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,2,power_law_1.01,0.5686592102050781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,2,power_law_1.01,0.3174463987350464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,2,power_law_1.01,0.5500736236572266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,2,power_law_1.01,0.3170815944671631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,2,power_law_1.01,0.5505023956298828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,2,power_law_1.01,0.31895039081573484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,2,power_law_1.01,0.5061952114105225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,2,power_law_1.01,0.5416895866394043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,2,power_law_1.01,0.3134016036987305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,2,balanced,14.019755045572916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,2,power_law_1.01,0.5593344211578369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,2,power_law_1.01,0.312224006652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,2,power_law_1.01,0.5857471942901611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,2,power_law_1.01,0.3349247932434082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,2,power_law_1.01,0.6162623882293701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,2,power_law_1.01,0.3630143880844116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,2,power_law_1.01,0.654201602935791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,2,power_law_1.01,0.40615038871765136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,2,power_law_1.01,0.6813375949859619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,2,power_law_1.01,0.43457918167114257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,2,power_law_1.01,0.9271360397338867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,2,power_law_1.01,1.105504035949707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,2,power_law_1.01,1.2392191886901855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,2,power_law_1.01,1.760371208190918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,2,power_law_1.01,2.1487167358398436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,2,power_law_1.01,2.9502847671508787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,2,power_law_1.01,4.046444702148437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,2,power_law_1.01,5.539846420288086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,2,power_law_1.01,7.900838470458984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,2,power_law_1.01,11.05121307373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,2,power_law_1.01,13.522802734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,2,power_law_1.01,18.09706268310547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,2,power_law_1.01,26.317279052734374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,2,power_law_1.01,50.99956359863281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,2,power_law_1.01,0.1602112054824829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,2,power_law_1.01,0.1777343988418579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,2,power_law_1.01,0.23877758979797364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,2,power_law_1.01,0.30001919269561766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,2,power_law_1.01,0.29948160648345945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,2,power_law_1.01,0.3115328073501587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,2,power_law_1.01,0.30662400722503663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,2,power_law_1.01,0.2985663890838623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,2,power_law_1.01,0.29818880558013916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,2,power_law_1.01,0.2930880069732666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,2,power_law_1.01,0.28908801078796387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,2,power_law_1.01,0.30743680000305174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,2,power_law_1.01,0.3151360034942627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,2,power_law_1.01,0.3208127975463867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,2,power_law_1.01,0.3652928113937378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,2,power_law_1.01,0.3819391965866089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,2,power_law_1.01,0.42842879295349123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,2,power_law_1.01,0.4690112113952637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.01,0.6074624061584473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.01,0.7636991977691651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.01,0.9333439826965332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.01,1.23437442779541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.01,1.5560895919799804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.01,2.500083160400391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.01,3.061996841430664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.01,4.359104156494141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.01,5.260332870483398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,2,power_law_1.01,6.750892639160156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,2,power_law_1.01,12.33565444946289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,2,power_law_1.01,0.04745599925518036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,2,power_law_1.01,0.050432002544403075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,2,power_law_1.01,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,2,power_law_1.01,25.606912231445314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,2,power_law_1.01,0.05898879766464234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,2,power_law_1.01,0.09123839735984803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,2,power_law_1.01,0.09014400243759155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,2,power_law_1.01,0.07107840180397033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,2,power_law_1.01,0.09335039854049683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,2,power_law_1.01,0.0889855980873108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,2,power_law_1.01,0.0780672013759613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,2,power_law_1.01,0.10094720125198364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,2,power_law_1.01,0.18122880458831786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,2,power_law_1.01,0.17874560356140137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,2,power_law_1.01,0.2163007974624634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,2,power_law_1.01,0.1450495958328247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,2,power_law_1.01,0.12133760452270508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,2,power_law_1.01,0.15870720148086548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,2,power_law_1.01,0.20565121173858641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.01,0.26652159690856936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.01,0.3734911918640137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.01,0.49351038932800295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.01,0.7058688163757324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.01,0.9311679840087891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.01,1.3707967758178712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.01,1.8019584655761718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.01,2.6807104110717774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.01,3.555833435058594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,2,power_law_1.2,0.2842047929763794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,2,power_law_1.01,4.433785629272461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,2,power_law_1.2,0.3799999952316284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,2,power_law_1.01,7.058592224121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,2,power_law_1.2,0.39258880615234376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,2,power_law_1.2,0.5874944210052491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,2,power_law_1.01,14.073536682128907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,2,power_law_1.2,0.5685056209564209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,2,power_law_1.2,0.5750976085662842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,2,power_law_1.2,0.5561471939086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,2,power_law_1.2,0.5665535926818848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,2,power_law_1.2,0.5466432094573974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,2,power_law_1.2,0.5544511795043945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,2,power_law_1.2,0.5191423892974854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,2,power_law_1.2,0.6115263938903809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,2,power_law_1.2,0.6085567951202393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,2,power_law_1.2,0.6643839836120605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,2,power_law_1.2,0.6924223899841309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,2,power_law_1.2,0.7478335857391357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,2,power_law_1.2,0.8046015739440918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,2,power_law_1.2,0.9397760391235351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,2,power_law_1.2,1.170304012298584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,2,power_law_1.2,1.9715520858764648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,2,power_law_1.2,2.4594047546386717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,2,power_law_1.2,3.2765823364257813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,2,power_law_1.01,0.45873279571533204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,2,power_law_1.01,0.573363208770752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,2,power_law_1.2,5.108019256591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,2,power_law_1.01,0.7374400138854981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,2,power_law_1.2,6.141350555419922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,2,power_law_1.01,0.9586624145507813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,2,power_law_1.2,7.7342979431152346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,2,power_law_1.01,1.287769603729248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,2,power_law_1.01,1.5491840362548828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,2,power_law_1.2,13.020903015136719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,2,power_law_1.01,1.995692825317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,2,power_law_1.01,2.725836753845215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,2,power_law_1.2,16.621414184570312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,2,power_law_1.01,4.303174209594727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,2,power_law_1.2,18.826393127441406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,2,power_law_1.01,4.807321548461914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,2,power_law_1.01,6.559993743896484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,2,power_law_1.2,27.730746459960937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,2,power_law_1.01,7.829209899902343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,2,power_law_1.01,14.110610961914062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,2,power_law_1.2,57.670849609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,2,power_law_1.01,29.99893798828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,2,power_law_1.2,0.17048319578170776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,2,power_law_1.2,0.18398720026016235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,2,power_law_1.2,0.2231231927871704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,2,power_law_1.2,0.3154623985290527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,2,power_law_1.2,0.3564800024032593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,2,power_law_1.2,0.3377984046936035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,2,power_law_1.2,0.3405184030532837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,2,power_law_1.2,0.3153023958206177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,2,power_law_1.2,0.3268800020217896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,2,power_law_1.2,0.3225663900375366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,2,power_law_1.2,0.31240320205688477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,2,power_law_1.2,0.3241152048110962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,2,power_law_1.2,0.36177279949188235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,2,power_law_1.2,0.34856319427490234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,2,power_law_1.2,0.398470401763916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,2,power_law_1.2,0.39580159187316893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,2,power_law_1.2,0.47712001800537107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,2,power_law_1.2,0.610591983795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,2,power_law_1.2,0.7262527942657471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,2,power_law_1.2,1.0010751724243163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,2,power_law_1.2,1.2835328102111816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,2,power_law_1.2,1.7155519485473634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,2,power_law_1.2,2.5281280517578124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,2,power_law_1.2,3.435955047607422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,2,power_law_1.2,4.098739242553711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,2,power_law_1.2,5.575372695922852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,2,power_law_1.2,7.6949821472167965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,2,power_law_1.2,0.1867967963218689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,2,power_law_1.2,0.17441920042037964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,2,power_law_1.2,9.472672271728516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,2,power_law_1.2,0.21339519023895265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,2,power_law_1.2,0.30616960525512693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,2,power_law_1.2,13.418540954589844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,2,power_law_1.2,0.2986176013946533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,2,power_law_1.2,0.3151808023452759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,2,power_law_1.2,0.2842623949050903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,2,power_law_1.2,26.760052490234376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,2,power_law_1.2,0.3079999923706055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,2,power_law_1.2,0.31058559417724607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,2,power_law_1.2,0.3010688066482544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,2,power_law_1.2,0.28574719429016116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,2,power_law_1.2,0.30634241104125975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,2,power_law_1.2,0.32717440128326414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,2,power_law_1.2,0.35123839378356936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,2,power_law_1.2,0.3652479887008667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,2,power_law_1.2,0.36830079555511475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,2,power_law_1.2,0.4254335880279541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,2,power_law_1.2,0.5133632183074951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.2,0.5710336208343506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.2,0.7744768142700196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.2,0.9478912353515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.2,1.2296128273010254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.2,1.8542783737182618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.2,2.034534454345703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.2,3.208710479736328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.2,4.228108978271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.2,5.764384078979492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,2,power_law_1.2,7.874022674560547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,2,power_law_1.2,10.684947204589843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,2,power_law_1.2,22.716709899902344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,2,power_law_1.2,0.04803839921951294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,1,balanced,0.1869279940923055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,2,power_law_1.2,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,1,balanced,0.30315732955932617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,2,power_law_1.2,0.0462336003780365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,1,balanced,0.5363786617914835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,2,power_law_1.2,0.05772160291671753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,1,balanced,0.510757327079773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,2,power_law_1.2,0.08960000276565552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,1,balanced,0.5611040194829305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,2,power_law_1.2,0.0865343987941742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,1,balanced,0.5422933499018351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,2,power_law_1.2,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,1,balanced,0.5355146725972494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,2,power_law_1.2,0.09176959991455078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,1,balanced,0.5413013299306234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,2,power_law_1.2,0.0883903980255127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,1,balanced,0.5119946797688802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,2,power_law_1.2,0.0766975998878479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,1,balanced,0.5060106515884399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,1,balanced,0.5452373425165812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,1,balanced,0.507477323214213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,1,balanced,0.5046613216400146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,1,balanced,0.47707200050354004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,1,balanced,0.44711466630299884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,1,balanced,0.43995734055836994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,1,balanced,0.44362131754557294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,1,balanced,0.7890133062998453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,1,balanced,0.7536853154500326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,1,balanced,1.0476640065511067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,1,balanced,1.3478719393412273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,1,balanced,1.9010826746622722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,1,balanced,2.405541261037191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,1,balanced,3.355050722757975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,1,balanced,4.282154719034831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,1,balanced,6.1209869384765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,1,balanced,0.11251733700434367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,1,balanced,0.20009599129358926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,1,balanced,8.052693049112955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,1,balanced,0.33690667152404785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,1,balanced,0.2985919912656148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,1,balanced,0.30056534210840863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,1,balanced,0.30035199721654254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,1,balanced,0.30403733253479004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,1,balanced,0.30112000306447345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,1,balanced,9.951103846232096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,1,balanced,0.330677330493927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,1,balanced,0.3057439923286438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,1,balanced,0.30727465947469074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,1,balanced,0.2763040065765381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,1,balanced,0.27106666564941406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,1,balanced,0.2685920000076294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,1,balanced,0.2651679913202922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,1,balanced,0.25784534215927124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,1,balanced,0.26654932896296185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,1,balanced,15.765205383300781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,1,balanced,0.45394666989644367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,1,balanced,0.4255839983622233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,1,balanced,0.5441279808680216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,1,balanced,0.6820586522420248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,1,balanced,0.9452746709187826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,1,balanced,1.2160639762878418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,1,balanced,1.7507039705912273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,1,balanced,2.2795680363972983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,1,balanced,31.763158162434895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,1,balanced,3.320927937825521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,1,balanced,0.12619200348854065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,1,balanced,0.20043732722600302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,1,balanced,0.34376001358032227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,1,balanced,4.391706784566243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,1,balanced,0.31699200471242267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,1,balanced,0.279530664285024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,1,balanced,0.3041973312695821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,1,balanced,0.3074880043665568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,1,balanced,0.2922293345133464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,1,balanced,5.511930465698242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,1,balanced,0.28224533796310425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,1,balanced,0.28597333033879596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,1,balanced,0.29532267649968463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,1,balanced,0.259061336517334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,1,balanced,0.2702239950497945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,1,balanced,0.2497439980506897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,1,balanced,8.752000172932943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,1,balanced,0.23337600628534952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,1,balanced,0.22320000330607095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,1,balanced,0.20896534125010172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,1,balanced,0.41873598098754883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,1,balanced,0.31933865944544476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,1,balanced,0.4273973306020101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,1,balanced,0.5457119941711426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,1,balanced,0.7741440137227377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,1,balanced,1.0030346711476643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,1,balanced,17.19326909383138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,1,balanced,1.4704000155131023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,1,balanced,1.9430826505025227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,1,balanced,2.8807252248128257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,1,balanced,3.8050400416056314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,1,balanced,4.717973391215007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,1,balanced,7.506133397420247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,2,power_law_1.2,0.09802240133285522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,2,power_law_1.2,0.177183997631073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,2,power_law_1.2,0.17960959672927856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,2,power_law_1.2,0.21687679290771483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,2,power_law_1.2,0.14982399940490723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,2,power_law_1.2,0.12174079418182374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,2,power_law_1.2,0.15808000564575195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,2,power_law_1.2,0.20521600246429444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.2,0.26506240367889405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.2,0.37279999256134033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.2,0.49095678329467773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.2,0.7054592132568359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,1,balanced,0.02796799937884013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,1,balanced,15.059200286865234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.2,0.9319295883178711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,1,balanced,0.029125332832336426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,1,balanced,0.028725333511829376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.2,1.3652095794677734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,1,balanced,0.03336533407370249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.2,1.80883846282959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,1,balanced,0.046767999728520714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.2,2.678047943115234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,1,balanced,0.06425066788991292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.2,3.556108856201172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,1,balanced,0.054234668612480164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,1,balanced,0.06333333253860474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,2,power_law_1.2,4.429151916503907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,1,balanced,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,2,power_law_1.2,7.044019317626953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,1,balanced,0.05676266551017761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,1,balanced,0.06653866668542226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,1,balanced,0.059706668059031166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,2,power_law_1.2,14.043661499023438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,1,balanced,0.06518933176994324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,1,balanced,0.07751999795436859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,1,balanced,0.15621333320935568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,1,balanced,0.17745065689086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,1,balanced,0.21150400241216025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,1,balanced,0.11763733625411987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,1,balanced,0.14070933063824972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,1,balanced,0.19541333119074503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,1,balanced,0.2457546591758728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,1,power_law_1.01,0.18466559648513795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,1,balanced,0.34648533662160236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,1,power_law_1.01,0.25153920650482176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,1,power_law_1.01,0.3011904001235962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,1,balanced,0.443178653717041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,1,power_law_1.01,0.5134848117828369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,1,power_law_1.01,0.49912319183349607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,1,balanced,0.6419999996821085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,1,power_law_1.01,0.5336832046508789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,1,power_law_1.01,0.5452672004699707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,1,power_law_1.01,0.5462592124938965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,1,balanced,0.8413386344909668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,1,power_law_1.01,0.5266111850738525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,1,power_law_1.01,0.5145472049713135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,1,power_law_1.01,0.5338751792907714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,1,power_law_1.01,0.5315072059631347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,1,balanced,1.2338933149973552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,1,power_law_1.01,0.5572800159454345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,1,power_law_1.01,0.5586112022399903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,1,power_law_1.01,0.6258944034576416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,1,power_law_1.01,0.6440320014953613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,1,power_law_1.01,0.6862527847290039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,1,balanced,1.6284054120381672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,1,power_law_1.01,0.8445631980895996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,1,power_law_1.01,0.9830207824707031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,1,power_law_1.01,1.2663040161132812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,1,power_law_1.01,1.5393471717834473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,1,balanced,2.022143999735514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,1,power_law_1.01,2.0184192657470703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,1,power_law_1.01,2.6079872131347654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,1,power_law_1.01,3.5179393768310545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,1,power_law_1.01,4.470828628540039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,1,power_law_1.01,6.39354248046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,1,balanced,3.2001546223958335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,1,power_law_1.01,8.182035064697265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,1,power_law_1.01,10.117874908447266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,1,power_law_1.01,15.943283081054688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,1,power_law_1.01,31.71962890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,1,balanced,6.348335901896159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,1,power_law_1.01,0.12748160362243652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,1,power_law_1.01,0.13712639808654786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,1,power_law_1.01,0.17827199697494506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,1,power_law_1.01,0.2974976062774658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,1,power_law_1.01,0.3002624034881592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,1,power_law_1.01,0.29270401000976565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,1,power_law_1.01,0.11461759805679321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,1,power_law_1.01,0.31626880168914795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,1,power_law_1.01,0.15406080484390258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,1,power_law_1.01,0.3165695905685425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,1,power_law_1.01,0.15214719772338867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,1,power_law_1.01,0.3081536054611206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,1,power_law_1.01,0.2913536071777344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,1,power_law_1.01,0.30467839241027833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,1,power_law_1.01,0.29747838973999025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,1,power_law_1.01,0.30357120037078855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,1,power_law_1.01,0.27936639785766604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,1,power_law_1.01,0.2902400016784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,1,power_law_1.01,0.2956543922424316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,1,power_law_1.01,0.29832959175109863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,1,power_law_1.01,0.30405120849609374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,1,power_law_1.01,0.2829632043838501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,1,power_law_1.01,0.3123647928237915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,1,power_law_1.01,0.2857088088989258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,1,power_law_1.01,0.34710400104522704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,1,power_law_1.01,0.2915647983551025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,1,power_law_1.01,0.3713792085647583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,1,power_law_1.01,0.26721279621124266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,1,power_law_1.01,0.39448320865631104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,1,power_law_1.01,0.30856320858001707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,1,power_law_1.01,0.4715263843536377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,1,power_law_1.01,0.31000959873199463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,1,power_law_1.01,0.5612544059753418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,1,power_law_1.01,0.31832959651947024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,1,power_law_1.01,0.6538432121276856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,1,power_law_1.01,0.3255743980407715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,1,power_law_1.01,0.8164863586425781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,1,power_law_1.01,0.355129599571228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,1,power_law_1.01,1.0974464416503906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,1,power_law_1.01,0.40952320098876954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,1,power_law_1.01,1.3425024032592774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.01,0.4804351806640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,1,power_law_1.01,1.8704448699951173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.01,0.5946815967559814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.01,0.7110591888427734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,1,power_law_1.01,2.426188850402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.01,0.9339200019836426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,1,power_law_1.01,3.500300979614258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.01,1.1730239868164063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,1,power_law_1.01,4.52344970703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.01,1.600876808166504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,1,power_law_1.01,5.620191955566407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.01,2.07127685546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.01,3.0023040771484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,1,power_law_1.01,8.720851135253906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.01,3.9124225616455077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,1,power_law_1.01,17.206118774414062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.01,4.854246520996094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,1,power_law_1.01,7.684806060791016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,1,power_law_1.01,15.326669311523437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,1,power_law_1.01,0.02619520127773285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,1,power_law_1.01,0.02815360128879547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,1,power_law_1.01,0.02826879918575287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,1,power_law_1.01,0.0320576012134552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,1,power_law_1.01,0.04338560104370117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,1,power_law_1.01,0.060057598352432254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,1,power_law_1.01,0.05103359818458557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,1,power_law_1.01,0.060678398609161376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,1,power_law_1.01,0.0506879985332489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,1,power_law_1.2,0.18858879804611206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,1,power_law_1.01,0.05516800284385681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,1,power_law_1.2,0.2330751895904541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,1,power_law_1.01,0.06358399987220764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,1,power_law_1.2,0.28764159679412843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,1,power_law_1.01,0.05927680134773254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,1,power_law_1.2,0.5162687778472901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,1,power_law_1.01,0.06354560256004334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,1,power_law_1.2,0.5156032085418701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,1,power_law_1.01,0.07390080094337463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,1,power_law_1.2,0.5355072021484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,1,power_law_1.01,0.15416959524154664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,1,power_law_1.2,0.5271615982055664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,1,power_law_1.01,0.1769536018371582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,1,power_law_1.2,0.5431935787200928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,1,power_law_1.01,0.2096384048461914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,1,power_law_1.2,0.5136896133422851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,1,power_law_1.01,0.11507840156555176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,1,power_law_1.2,0.5160639762878418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.01,0.1395583987236023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,1,power_law_1.2,0.5158271789550781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,1,power_law_1.2,0.5456319808959961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.01,0.19317760467529296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.01,0.243833589553833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,1,power_law_1.2,0.5658048152923584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,1,power_law_1.2,0.5867008209228516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.01,0.34535040855407717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,1,power_law_1.2,0.636249589920044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.01,0.4422272205352783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,1,power_law_1.2,0.6102848052978516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.01,0.6429056167602539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,1,power_law_1.2,0.7077760219573974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.01,0.8378239631652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,1,power_law_1.2,0.8790783882141113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.01,1.235929584503174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,1,power_law_1.2,0.9848192214965821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.01,1.6308160781860352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,1,power_law_1.2,1.2714240074157714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.01,2.0213375091552734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,1,power_law_1.2,1.4690303802490234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,1,power_law_1.01,3.20568962097168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,1,power_law_1.2,2.071455955505371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,1,power_law_1.01,6.358227157592774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,1,power_law_1.2,2.558720016479492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,1,power_law_1.2,3.5933441162109374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,1,power_law_1.2,4.48600959777832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,1,power_law_1.2,6.315846252441406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,1,power_law_1.2,8.21107177734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,1,power_law_1.2,10.10816650390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,1,power_law_1.2,15.955354309082031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,1,power_law_1.2,31.588473510742187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,1,power_law_1.2,0.11005439758300781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,1,power_law_1.2,0.1123263955116272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,1,power_law_1.2,0.14214400053024293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,1,power_law_1.2,0.13782399892807007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,1,power_law_1.2,0.18045439720153808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,1,power_law_1.2,0.19890559911727906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,1,power_law_1.2,0.28064000606536865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,1,power_law_1.2,0.2889343976974487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,1,power_law_1.2,0.3041088104248047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,1,power_law_1.2,0.2757823944091797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,1,power_law_1.2,0.30347518920898436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,1,power_law_1.2,0.2991231918334961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,1,power_law_1.2,0.28881919384002686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,1,power_law_1.2,0.2936511993408203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,1,power_law_1.2,0.28616321086883545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,1,power_law_1.2,0.30344319343566895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,1,power_law_1.2,0.2943360090255737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,1,power_law_1.2,0.2976320028305054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,1,power_law_1.2,0.2750272035598755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,1,power_law_1.2,0.303219199180603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,1,power_law_1.2,0.28090879917144773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,1,power_law_1.2,0.3283519983291626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,1,power_law_1.2,0.2834111928939819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,1,power_law_1.2,0.3199295997619629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,1,power_law_1.2,0.2828991889953613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,1,power_law_1.2,0.33162879943847656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,1,power_law_1.2,0.2869247913360596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,1,power_law_1.2,0.3536191940307617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,1,power_law_1.2,0.28803839683532717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,1,power_law_1.2,0.368889594078064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,1,power_law_1.2,0.33613440990447996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,1,power_law_1.2,0.4059391975402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,1,power_law_1.2,0.3516223907470703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,1,power_law_1.2,0.4757887840270996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,1,power_law_1.2,0.354041600227356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,1,power_law_1.2,0.5299712181091308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,1,power_law_1.2,0.43710718154907224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,1,power_law_1.2,0.6973184108734131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.2,0.4581888198852539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.2,0.5872128009796143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,1,power_law_1.2,0.8323519706726075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.2,0.6760255813598632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,1,power_law_1.2,1.096019172668457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.2,0.9042176246643067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,1,power_law_1.2,1.334432029724121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.2,1.1383232116699218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,1,power_law_1.2,1.8902847290039062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.2,1.6114559173583984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,1,power_law_1.2,2.388607978820801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.2,2.050815963745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,1,power_law_1.2,3.5065025329589843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.2,3.0058752059936524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,1,power_law_1.2,4.506931304931641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.2,3.896307373046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,1,power_law_1.2,5.620390319824219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.2,4.833472061157226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,1,power_law_1.2,8.66845474243164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,1,power_law_1.2,7.617279815673828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,1,power_law_1.2,17.339532470703126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,1,power_law_1.2,15.197433471679688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,1,power_law_1.2,0.0271232008934021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,1,power_law_1.2,0.0286655992269516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,1,power_law_1.2,0.027564799785614012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,1,power_law_1.2,0.031513598561286923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,1,power_law_1.2,0.043289598822593686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,1,power_law_1.2,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,1,power_law_1.2,0.05030400156974792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,1,power_law_1.2,0.05912320017814636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,1,power_law_1.2,0.0513152003288269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,1,power_law_1.2,0.05459840297698974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,1,power_law_1.2,0.0637440025806427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,1,power_law_1.2,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,1,power_law_1.2,0.06209920048713684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,1,power_law_1.2,0.07470080256462097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,1,power_law_1.2,0.1542464017868042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,1,power_law_1.2,0.17704319953918457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,1,power_law_1.2,0.2102976083755493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,1,power_law_1.2,0.11439360380172729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.2,0.1399616003036499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.2,0.1955072045326233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.2,0.2443392038345337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.2,0.34591360092163087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.2,0.44353280067443845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.2,0.6427648067474365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.2,0.8389504432678223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.2,1.234995174407959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.2,1.6273216247558593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.2,2.016979217529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,1,power_law_1.2,3.2071422576904296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,1,power_law_1.2,6.361350250244141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,2,balanced,0.1304746667544047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,2,balanced,0.20012267430623373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,2,balanced,0.33532265822092694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,2,balanced,0.5687946478525797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,2,balanced,1.013856013615926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,2,balanced,1.8459307352701824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,2,balanced,1.8493173917134602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,2,balanced,1.8998239835103352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,2,balanced,1.8699307441711426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,2,balanced,1.9776533444722493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,2,balanced,1.8625067075093586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,2,balanced,1.8835840225219727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,2,balanced,2.00492270787557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,2,balanced,1.8232639630635579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,2,balanced,1.9519999821980794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,2,balanced,1.9001599947611492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,2,balanced,1.8698347409566243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,2,balanced,1.885200023651123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,2,balanced,1.918384075164795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,2,balanced,1.9040746688842773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,2,balanced,1.9242560068766277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,2,balanced,1.9705653190612793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,2,balanced,2.0573013623555503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,2,balanced,0.0956213374932607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,2,balanced,0.13285332918167114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,2,balanced,0.19913599888483682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,2,balanced,3.300133387247721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,2,balanced,0.34004799524943036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,2,balanced,0.557914654413859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,2,balanced,0.10156800349553426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,2,balanced,1.010805368423462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,2,balanced,0.14080533385276794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,2,balanced,1.0071093241373699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,2,balanced,0.20433600743611655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,2,balanced,0.9704853693644205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,2,balanced,0.3516639868418376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,2,balanced,0.9786293506622314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,2,balanced,3.086117426554362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,2,balanced,0.551962653795878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,2,balanced,0.982373317082723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,2,balanced,1.0283253192901611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,2,balanced,1.0076693693796794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,2,balanced,0.9652960300445557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,2,balanced,0.9880053202311198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,2,balanced,0.9178240299224854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,2,balanced,1.0423519611358643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,2,balanced,0.9370559851328532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,2,balanced,1.0392213662465413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,2,balanced,0.9089972972869873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,2,balanced,1.0093759695688884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,2,balanced,0.9074719746907552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,2,balanced,1.1023680369059246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,2,balanced,0.9092960357666016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,2,balanced,1.0159306526184082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,2,balanced,0.9016053676605225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,2,balanced,4.259610811869304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,2,balanced,0.9334346453348795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,2,balanced,1.0823733011881511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,2,balanced,0.9458080132802328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,2,balanced,1.0545492966969807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,2,balanced,1.002410650253296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,2,balanced,0.976576010386149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,2,balanced,1.085103988647461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,2,balanced,1.0243199666341145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,2,balanced,1.106874704360962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,2,balanced,0.9699146747589111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,2,balanced,1.010602633158366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,2,balanced,1.196122646331787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,2,balanced,5.524890899658203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,2,balanced,0.9904906749725342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,2,balanced,1.2967519760131836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,2,balanced,1.052506685256958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,2,balanced,1.130405346552531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,2,balanced,2.0529653231302896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,2,balanced,1.9207199414571126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,2,balanced,0.050069332122802734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,2,balanced,6.798159917195638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,2,balanced,0.06365866462389629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,2,balanced,0.09303999940554301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,2,balanced,2.014570713043213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,2,balanced,0.14286933342615762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,2,balanced,0.2493706742922465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,2,balanced,1.911861260732015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,2,balanced,0.45313068230946857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,2,balanced,0.46692268053690594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,2,balanced,0.47678931554158527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,2,balanced,0.48901331424713135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,2,balanced,0.4997653166453044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,2,balanced,0.523141344388326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,2,balanced,0.5000053246816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,2,balanced,0.5102186600367228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,2,balanced,2.6545119285583496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,2,balanced,0.5321439901987711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,2,balanced,2.816570599873861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,2,balanced,0.5451733271280924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,2,balanced,0.5464479923248291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,2,balanced,0.5570559899012247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,2,balanced,0.7366399765014648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,2,balanced,0.7046026388804117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,2,balanced,1.2906453609466553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,2,balanced,3.441125233968099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,2,balanced,10.780741373697916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,2,balanced,1.4085173606872559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,2,balanced,3.6328372955322266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,2,balanced,1.1354986826578777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,2,balanced,1.1747679710388184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,2,balanced,1.2499573230743408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,2,balanced,4.257295926411946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,2,balanced,1.3298933506011963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,2,balanced,4.469936052958171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,2,balanced,2.358250617980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,2,balanced,6.682975769042969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,2,balanced,2.4172693888346353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,2,balanced,7.064554850260417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,2,balanced,21.78039042154948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,2,balanced,3.48525333404541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,2,balanced,13.140431722005209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,2,balanced,4.633573214213054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,2,balanced,14.447060902913412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,2,balanced,9.15716298421224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,2,power_law_1.01,0.19494400024414063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,2,power_law_1.01,0.3530495882034302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,2,power_law_1.01,0.2798847913742065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,2,power_law_1.01,0.4641024112701416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,2,power_law_1.01,0.6924287796020507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,2,power_law_1.01,0.96693115234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,2,power_law_1.01,1.3421119689941405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,2,power_law_1.01,1.4296192169189452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,2,power_law_1.01,1.4291263580322267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,2,power_law_1.01,1.5158399581909179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,2,power_law_1.01,1.5097344398498536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,2,power_law_1.01,1.5881983757019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,2,power_law_1.01,1.6622528076171874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,2,power_law_1.01,1.6338943481445312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,2,power_law_1.01,1.663430404663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,2,power_law_1.01,0.13176319599151612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,2,power_law_1.01,0.2042623996734619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,2,power_law_1.01,1.7287679672241212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,2,power_law_1.01,0.1463039994239807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,2,power_law_1.01,1.7201215744018554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,2,power_law_1.01,0.25852160453796386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,2,power_law_1.01,1.7598911285400392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,2,power_law_1.01,0.4055232048034668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,2,power_law_1.01,1.8815807342529296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,2,power_law_1.01,0.5270400047302246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,2,power_law_1.01,2.193497657775879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,2,power_law_1.01,0.7008959770202636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,2,power_law_1.01,2.3007232666015627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,2,power_law_1.01,0.7516223907470703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,2,power_law_1.01,2.727916717529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,2,power_law_1.01,0.7869247913360595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,2,power_law_1.01,3.0922048568725584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,2,power_law_1.01,0.8266176223754883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,2,power_law_1.01,4.021945571899414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,2,power_law_1.01,0.8155136108398438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,2,power_law_1.01,4.60761604309082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,2,power_law_1.01,0.8499072074890137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,2,power_law_1.01,6.117510223388672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,2,power_law_1.01,0.8683456420898438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,2,power_law_1.01,0.8934016227722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,2,power_law_1.01,7.915116882324218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,2,power_law_1.01,0.9202431678771973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,2,power_law_1.01,9.279078674316406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,2,power_law_1.01,0.9329919815063477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,2,power_law_1.01,0.9352959632873535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,2,power_law_1.01,12.908274841308593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,2,power_law_1.01,0.9913472175598145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,2,power_law_1.01,0.15564800500869752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,2,power_law_1.01,1.0436479568481445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,2,power_law_1.01,0.21457281112670898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,2,power_law_1.01,1.1971072196960448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,2,power_law_1.01,26.12359619140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,2,power_law_1.01,0.17052160501480101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,2,power_law_1.01,0.31122560501098634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,2,power_law_1.01,1.3012800216674805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,2,power_law_1.01,0.43363199234008787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,2,power_law_1.01,1.55665283203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,2,power_law_1.01,0.553107213973999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,2,power_law_1.01,1.7981760025024414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,2,power_law_1.01,0.7194240093231201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,2,power_law_1.01,2.260211181640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,2,power_law_1.01,0.7814527988433838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,2,power_law_1.01,2.7306047439575196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,2,power_law_1.01,0.806764793395996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,2,power_law_1.01,3.6289791107177733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,2,power_law_1.01,0.8040512084960938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,2,power_law_1.01,4.739648056030274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,2,power_law_1.01,0.8385087966918945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,2,power_law_1.01,0.845740795135498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,2,power_law_1.01,5.550796890258789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,2,power_law_1.01,0.86428804397583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,2,power_law_1.01,8.112882995605469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,2,power_law_1.01,0.9020671844482422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,2,power_law_1.01,0.048511999845504764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,2,power_law_1.01,0.8978688240051269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,2,power_law_1.01,0.0762112021446228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,2,power_law_1.01,16.5578369140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,2,power_law_1.01,0.9136384010314942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,2,power_law_1.01,0.061286401748657224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,2,power_law_1.01,0.11315200328826905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,2,power_law_1.01,0.895628833770752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,2,power_law_1.01,0.16540160179138183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,2,power_law_1.01,0.9621120452880859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,2,power_law_1.01,0.23029758930206298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.01,1.0310208320617675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,2,power_law_1.01,0.31182079315185546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.01,1.122323226928711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,2,power_law_1.01,0.3477952003479004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.01,1.2422656059265136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,2,power_law_1.01,0.34882559776306155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.01,1.516806411743164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,2,power_law_1.01,0.3706687927246094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.01,1.7374719619750976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,2,power_law_1.01,0.3918463945388794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.01,2.294521522521973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,2,power_law_1.01,0.37557759284973147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.01,2.643935966491699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,2,power_law_1.01,0.3926527976989746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.01,3.7506622314453124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,2,power_law_1.01,0.4090112209320068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,2,power_law_1.01,0.4144896030426025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.01,4.608147048950196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,2,power_law_1.01,0.4294015884399414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,2,power_law_1.01,5.37336311340332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,2,power_law_1.01,0.45525121688842773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,2,power_law_1.01,0.49319682121276853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,2,power_law_1.01,8.077811431884765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.01,0.5019519805908204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.01,1.106816005706787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,2,power_law_1.01,15.079866027832031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.01,1.25665283203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.01,0.8350399971008301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.01,0.9301504135131836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.01,1.1277440071105957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.01,1.2889408111572265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.01,1.6861440658569335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.01,2.0425920486450195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,2,power_law_1.01,2.5073856353759765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,2,power_law_1.01,3.7451839447021484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,2,power_law_1.01,7.299372863769531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,2,power_law_1.2,0.1995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,2,power_law_1.2,0.33374719619750975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,2,power_law_1.2,0.22266879081726074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,2,power_law_1.2,0.395795202255249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,2,power_law_1.2,0.5893055915832519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,2,power_law_1.2,0.8438272476196289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,2,power_law_1.2,1.2953472137451172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,2,power_law_1.2,1.3058815956115724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,2,power_law_1.2,1.435807991027832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,2,power_law_1.2,1.5001215934753418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,2,power_law_1.2,1.463590431213379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,2,power_law_1.2,1.590227222442627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,2,power_law_1.2,1.6510911941528321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,2,power_law_1.2,0.1307136058807373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,2,power_law_1.2,1.6457855224609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,2,power_law_1.2,0.1926911950111389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,2,power_law_1.2,1.687833595275879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,2,power_law_1.2,0.1579967975616455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,2,power_law_1.2,1.756902313232422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,2,power_law_1.2,0.2600831985473633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,2,power_law_1.2,0.35477120876312257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,2,power_law_1.2,1.826348876953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,2,power_law_1.2,0.46895999908447267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,2,power_law_1.2,1.9393535614013673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,2,power_law_1.2,0.6969791889190674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,2,power_law_1.2,2.0655935287475584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,2,power_law_1.2,0.7547455787658691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,2,power_law_1.2,2.2745088577270507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,2,power_law_1.2,0.7579135894775391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,2,power_law_1.2,2.4862848281860352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,2,power_law_1.2,0.7741375923156738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,2,power_law_1.2,2.9826496124267576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,2,power_law_1.2,0.8064767837524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,2,power_law_1.2,3.2521408081054686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,2,power_law_1.2,0.8605695724487304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,2,power_law_1.2,4.048844909667968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,2,power_law_1.2,0.8834624290466309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,2,power_law_1.2,0.9017600059509278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,2,power_law_1.2,4.711468887329102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,2,power_law_1.2,0.9346495628356933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,2,power_law_1.2,6.272895812988281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,2,power_law_1.2,0.949407958984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,2,power_law_1.2,7.642342376708984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,2,power_law_1.2,0.9824383735656739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,2,power_law_1.2,1.0525952339172364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,2,power_law_1.2,9.308198547363281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,2,power_law_1.2,1.1268223762512206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,2,power_law_1.2,0.15411839485168458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,2,power_law_1.2,1.2354111671447754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,2,power_law_1.2,14.200845336914062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,2,power_law_1.2,0.22272639274597167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,2,power_law_1.2,1.3492735862731933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,2,power_law_1.2,0.14807039499282837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,2,power_law_1.2,1.6612287521362306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,2,power_law_1.2,0.29068799018859864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,2,power_law_1.2,26.931295776367186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,2,power_law_1.2,1.8400127410888671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,2,power_law_1.2,0.3617919921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,2,power_law_1.2,2.3460927963256837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,2,power_law_1.2,0.5187392234802246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,2,power_law_1.2,0.7193088054656982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,2,power_law_1.2,2.79370231628418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,2,power_law_1.2,0.7578815937042236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,2,power_law_1.2,3.8003841400146485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,2,power_law_1.2,0.7672383785247803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,2,power_law_1.2,5.080422210693359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,2,power_law_1.2,0.8149632453918457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,2,power_law_1.2,0.04821119904518127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,2,power_law_1.2,5.9412799835205075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,2,power_law_1.2,0.8524224281311035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,2,power_law_1.2,0.06993280053138733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,2,power_law_1.2,0.8779775619506835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,2,power_law_1.2,0.06541439890861511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,2,power_law_1.2,8.902623748779297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,2,power_law_1.2,0.8736512184143066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,2,power_law_1.2,0.10938880443572999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,2,power_law_1.2,0.908409595489502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,2,power_law_1.2,0.14855680465698243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,2,power_law_1.2,18.256678771972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,2,power_law_1.2,0.9331328392028808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,2,power_law_1.2,0.2111680030822754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,2,power_law_1.2,0.9507136344909668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,2,power_law_1.2,0.3084480047225952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,2,power_law_1.2,0.33207681179046633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,2,power_law_1.2,0.9989888191223144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,2,power_law_1.2,0.3553663969039917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,2,power_law_1.2,1.0215167999267578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,2,power_law_1.2,0.36640000343322754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.2,1.116044807434082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,2,power_law_1.2,0.39390718936920166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.2,1.2267519950866699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,2,power_law_1.2,0.39000959396362306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.2,1.3433088302612304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,2,power_law_1.2,0.4089983940124512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.2,1.5522047996520996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,2,power_law_1.2,0.4257023811340332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.2,1.7974912643432617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,2,power_law_1.2,0.427891206741333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.2,2.2971519470214843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,2,power_law_1.2,0.42824320793151854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.2,2.744576072692871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,2,power_law_1.2,0.4699903964996338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.2,3.591999816894531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,2,power_law_1.2,0.526854419708252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.2,0.5341887950897217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.2,4.8898368835449215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.2,1.2058303833007813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,2,power_law_1.2,5.929049682617188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.2,1.3566271781921386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,2,power_law_1.2,8.28431396484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.2,0.9067328453063965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.2,1.029599952697754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,2,power_law_1.2,15.813702392578126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.2,1.1823360443115234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.2,1.3152576446533204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.2,1.8137216567993164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.2,2.259238433837891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,2,power_law_1.2,2.7047296524047852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,2,power_law_1.2,3.998489761352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,2,power_law_1.2,7.796256256103516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,1,balanced,0.13019200166066489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,1,balanced,0.19686400890350342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,1,balanced,0.32320000727971393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,1,balanced,0.5324853261311849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,1,balanced,0.9747093518575033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,1,balanced,1.816266695658366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,1,balanced,1.8091732660929363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,1,balanced,1.815125306447347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,1,balanced,1.8329013188680012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,1,balanced,1.855130672454834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,1,balanced,1.8456106185913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,1,balanced,0.09682133793830872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,1,balanced,1.841584046681722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,1,balanced,0.13467199603716531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,1,balanced,1.894063949584961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,1,balanced,0.20881599187850952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,1,balanced,1.8422187169392903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,1,balanced,0.33444801966349286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,1,balanced,0.5608959992726644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,1,balanced,1.925930658976237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,1,balanced,0.9961439768473307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,1,balanced,1.881824016571045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,1,balanced,0.9902986685434977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,1,balanced,1.8337333997090657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,1,balanced,0.9581973552703857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,1,balanced,0.9676160017649332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,1,balanced,1.9554559389750164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,1,balanced,0.9607946872711182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,1,balanced,0.976090669631958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,1,balanced,1.92192538579305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,1,balanced,0.966970682144165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,1,balanced,0.9985600312550863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,1,balanced,1.9817867279052734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,1,balanced,0.9830079873402914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,1,balanced,0.9953280289967855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,1,balanced,2.011258602142334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,1,balanced,1.0120213031768799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,1,balanced,1.0109492937723796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,1,balanced,1.0794666608174641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,1,balanced,2.1418026288350425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,1,balanced,1.092624028523763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,1,balanced,0.12814399600028992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,1,balanced,1.166981299718221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,1,balanced,0.16674667596817017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,1,balanced,2.278480052947998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,1,balanced,0.2288480003674825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,1,balanced,0.3685493469238281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,1,balanced,1.2309599717458088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,1,balanced,0.5911786556243896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,1,balanced,1.0786933104197185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,1,balanced,1.0461013317108154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,1,balanced,1.000912030537923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,1,balanced,3.54205322265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,1,balanced,1.3979786237080891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,1,balanced,0.9935359954833984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,1,balanced,0.9616373380025228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,1,balanced,0.9572479724884033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,1,balanced,0.9805440107981364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,1,balanced,1.5655412673950195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,1,balanced,0.9787786801656088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,1,balanced,0.9989226659138998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,1,balanced,3.459226608276367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,1,balanced,0.02940800040960312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,1,balanced,0.9817012945810953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,1,balanced,0.04484800000985464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,1,balanced,1.0209173361460369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,1,balanced,0.05939733485380808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,1,balanced,0.08540800213813782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,1,balanced,1.0295946598052979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,1,balanced,2.3091252644856772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,1,balanced,0.14095466335614523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,1,balanced,0.25235732396443683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,1,balanced,1.065557320912679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,1,balanced,0.25975465774536133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,1,balanced,0.2653973301251729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,1,balanced,1.0637280146280925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,1,balanced,0.27299733956654865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,1,balanced,4.845871925354004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,1,balanced,0.27931199471155804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,1,balanced,1.1341493129730225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,1,balanced,0.2935626705487569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,1,balanced,0.30664000908533734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,1,balanced,2.5419626235961914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,1,balanced,1.1511999766031902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,1,balanced,0.3213813304901123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,1,balanced,0.34909331798553467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,1,balanced,0.31383466720581055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,1,balanced,1.3000373045603435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,1,balanced,0.32898666461308795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,1,balanced,0.3553119897842407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,1,balanced,0.34842665990193683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,1,balanced,1.4730186462402344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,1,balanced,0.38159998257954914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,1,balanced,6.1734771728515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,1,balanced,3.524911880493164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,1,balanced,0.38978668053944904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,1,balanced,2.344330628712972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,1,balanced,0.42531200249989826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,1,balanced,0.8969866434733073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,1,balanced,2.542367935180664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,1,balanced,1.0337066650390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,1,balanced,4.62337080637614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,1,balanced,7.526682535807292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,1,balanced,0.7378079891204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,1,balanced,3.592522621154785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,1,balanced,0.8082079887390137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,1,balanced,5.727466583251953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,1,balanced,4.698224067687988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,1,balanced,1.3611680666605632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,1,balanced,12.285483042399088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,1,balanced,1.4827040036519368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,1,balanced,5.789909362792969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,1,balanced,9.22426668802897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,1,balanced,2.040991942087809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,1,balanced,9.20307731628418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,1,balanced,2.848714510599772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,1,balanced,25.102928161621094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,1,balanced,18.656517028808594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,1,balanced,17.86040496826172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,1,balanced,5.595984141031901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,1,power_law_1.01,0.12702080011367797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,1,power_law_1.01,0.19612159729003906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,1,power_law_1.01,0.09450240135192871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,1,power_law_1.01,0.23960320949554442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,1,power_law_1.01,0.12851840257644653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,1,power_law_1.01,0.4266176223754883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,1,power_law_1.01,0.14229120016098024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,1,power_law_1.01,0.6456128120422363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,1,power_law_1.01,0.2631680011749268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,1,power_law_1.01,0.8849791526794434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,1,power_law_1.01,0.38648960590362547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,1,power_law_1.01,1.2535360336303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,1,power_law_1.01,0.49407358169555665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,1,power_law_1.01,1.364243221282959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,1,power_law_1.01,0.6896959781646729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,1,power_law_1.01,1.3412159919738769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,1,power_law_1.01,0.7273536205291748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,1,power_law_1.01,0.756883192062378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,1,power_law_1.01,1.4620863914489746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,1,power_law_1.01,0.772812795639038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,1,power_law_1.01,1.4632767677307128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,1,power_law_1.01,0.8019904136657715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,1,power_law_1.01,1.5356672286987305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,1,power_law_1.01,0.8232959747314453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,1,power_law_1.01,1.5700927734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,1,power_law_1.01,0.860313606262207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,1,power_law_1.01,1.5892160415649415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,1,power_law_1.01,0.873964786529541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,1,power_law_1.01,1.644371223449707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,1,power_law_1.01,0.8989439964294433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,1,power_law_1.01,1.6490623474121093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,1,power_law_1.01,0.8968255996704102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,1,power_law_1.01,1.7011199951171876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,1,power_law_1.01,0.027628800272941588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,1,power_law_1.01,0.9409919738769531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,1,power_law_1.01,1.7606016159057618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,1,power_law_1.01,0.0502016007900238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,1,power_law_1.01,1.016153621673584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,1,power_law_1.01,1.910086441040039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,1,power_law_1.01,0.12501120567321777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,1,power_law_1.01,0.05081599950790405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,1,power_law_1.01,1.0854592323303223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,1,power_law_1.01,0.16010240316390992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,1,power_law_1.01,2.100223922729492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,1,power_law_1.01,0.07639679908752442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,1,power_law_1.01,1.2342656135559082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,1,power_law_1.01,0.1613312005996704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,1,power_law_1.01,0.13610880374908446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,1,power_law_1.01,2.326406478881836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,1,power_law_1.01,1.360307216644287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,1,power_law_1.01,0.30032639503479003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,1,power_law_1.01,0.1941375970840454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,1,power_law_1.01,2.691641616821289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,1,power_law_1.01,0.43144960403442384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,1,power_law_1.01,0.266483211517334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,1,power_law_1.01,1.6142208099365234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,1,power_law_1.01,3.1121599197387697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,1,power_law_1.01,0.30302720069885253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,1,power_law_1.01,0.5723648071289062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,1,power_law_1.01,1.9201536178588867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,1,power_law_1.01,3.8978816986083986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,1,power_law_1.01,0.3230720043182373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,1,power_law_1.01,0.739302396774292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,1,power_law_1.01,2.4539840698242186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,1,power_law_1.01,4.722886276245117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,1,power_law_1.01,0.32006399631500243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,1,power_law_1.01,0.7647808074951172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,1,power_law_1.01,3.0351680755615233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,1,power_law_1.01,0.3437824010848999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,1,power_law_1.01,0.8038975715637207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,1,power_law_1.01,6.280499267578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,1,power_law_1.01,4.145043182373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,1,power_law_1.01,0.36681599617004396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,1,power_law_1.01,0.8005311965942383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,1,power_law_1.01,7.823641967773438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,1,power_law_1.01,5.242399978637695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,1,power_law_1.01,0.3769151926040649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,1,power_law_1.01,0.8421631813049316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,1,power_law_1.01,0.406931209564209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,1,power_law_1.01,0.8465727806091309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,1,power_law_1.01,6.3274688720703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,1,power_law_1.01,9.357113647460938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,1,power_law_1.01,0.38782079219818116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,1,power_law_1.01,0.8728256225585938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,1,power_law_1.01,9.582592010498047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,1,power_law_1.01,0.40083842277526854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,1,power_law_1.01,14.179731750488282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,1,power_law_1.01,0.8912192344665527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,1,power_law_1.01,0.4403264045715332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,1,power_law_1.01,0.9010368347167969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,1,power_law_1.01,0.4313024044036865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,1,power_law_1.01,18.766323852539063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,1,power_law_1.01,0.9115391731262207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,1,power_law_1.01,27.15916748046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.01,0.47017598152160645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,1,power_law_1.01,0.9326656341552735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.01,0.5116735935211182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,1,power_law_1.01,1.0117823600769043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.01,0.5467391967773437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.01,1.086732769012451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.01,1.2972096443176269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.01,1.2253631591796874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.01,1.473094367980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.01,1.3602944374084474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.01,0.9431103706359864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.01,1.6298240661621093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.01,1.1264703750610352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.01,1.9167488098144532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.01,1.4515392303466796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.01,2.4780031204223634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.01,1.82872314453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.01,3.0466239929199217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.01,2.1396480560302735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.01,4.164409637451172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,1,power_law_1.01,3.2949249267578127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.01,5.2539520263671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,1,power_law_1.01,5.926764678955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.01,6.398105621337891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,1,power_law_1.01,9.695270538330078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,1,power_law_1.01,18.291526794433594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,1,power_law_1.2,0.1268928050994873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,1,power_law_1.2,0.1916607975959778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,1,power_law_1.2,0.09285119771957398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,1,power_law_1.2,0.21203200817108153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,1,power_law_1.2,0.12707200050354003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,1,power_law_1.2,0.3916735887527466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,1,power_law_1.2,0.1383679986000061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,1,power_law_1.2,0.596947193145752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,1,power_law_1.2,0.24684159755706786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,1,power_law_1.2,0.7631423950195313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,1,power_law_1.2,0.359769606590271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,1,power_law_1.2,1.2005887985229493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,1,power_law_1.2,0.4451648235321045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,1,power_law_1.2,1.2952128410339356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,1,power_law_1.2,0.6838655948638916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,1,power_law_1.2,1.3623744010925294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,1,power_law_1.2,0.7211840152740479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,1,power_law_1.2,1.3873536109924316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,1,power_law_1.2,1.4952704429626464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,1,power_law_1.2,0.7729023933410645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,1,power_law_1.2,1.5111295700073242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,1,power_law_1.2,0.7635519981384278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,1,power_law_1.2,1.5676032066345216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,1,power_law_1.2,0.7922688007354737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,1,power_law_1.2,1.5647551536560058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,1,power_law_1.2,0.8331839561462402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,1,power_law_1.2,0.11856640577316284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,1,power_law_1.2,0.8753984451293946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,1,power_law_1.2,1.656915283203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,1,power_law_1.2,0.1575935959815979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,1,power_law_1.2,0.9032320022583008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,1,power_law_1.2,1.6955392837524415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,1,power_law_1.2,0.17038079500198364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,1,power_law_1.2,0.9157376289367676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,1,power_law_1.2,1.741049575805664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,1,power_law_1.2,0.2664639949798584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,1,power_law_1.2,0.9399935722351074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,1,power_law_1.2,1.8810432434082032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,1,power_law_1.2,0.030291199684143066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,1,power_law_1.2,0.3931583881378174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,1,power_law_1.2,0.9967552185058594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,1,power_law_1.2,1.9973888397216797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,1,power_law_1.2,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,1,power_law_1.2,0.5163328170776367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,1,power_law_1.2,1.094591999053955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,1,power_law_1.2,2.171494483947754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,1,power_law_1.2,0.05012480020523071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,1,power_law_1.2,0.7205887794494629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,1,power_law_1.2,1.1363583564758302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,1,power_law_1.2,2.3739200592041017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,1,power_law_1.2,0.07960320115089417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,1,power_law_1.2,0.786348819732666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,1,power_law_1.2,1.3157183647155761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,1,power_law_1.2,0.12130559682846069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,1,power_law_1.2,2.789945602416992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,1,power_law_1.2,0.802393627166748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,1,power_law_1.2,0.8002047538757324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,1,power_law_1.2,0.17912319898605347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,1,power_law_1.2,3.14597110748291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,1,power_law_1.2,0.8199935913085937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,1,power_law_1.2,0.28067200183868407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,1,power_law_1.2,3.9694271087646484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,1,power_law_1.2,0.876863956451416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,1,power_law_1.2,0.2977535963058472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,1,power_law_1.2,4.748147201538086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,1,power_law_1.2,0.8873600006103516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,1,power_law_1.2,0.31519360542297364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,1,power_law_1.2,6.26115837097168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,1,power_law_1.2,0.9299712181091309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,1,power_law_1.2,0.31640961170196535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,1,power_law_1.2,0.9428863525390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,1,power_law_1.2,7.7460479736328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,1,power_law_1.2,0.3467200040817261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,1,power_law_1.2,0.36316800117492676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,1,power_law_1.2,0.9545663833618164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,1,power_law_1.2,9.313926696777344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,1,power_law_1.2,0.3804480075836182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,1,power_law_1.2,1.006662368774414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,1,power_law_1.2,1.0888383865356446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,1,power_law_1.2,0.40103678703308104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,1,power_law_1.2,14.075289916992187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.2,1.1581760406494142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,1,power_law_1.2,0.4014336109161377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.2,1.3055744171142578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,1,power_law_1.2,0.4044032096862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,1,power_law_1.2,26.323052978515626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,1,power_law_1.2,0.45854721069335935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.2,1.4299263954162598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,1,power_law_1.2,0.44795517921447753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.2,1.7197952270507812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.2,0.5066815853118897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.2,1.9983488082885743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.2,0.5143936157226563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.2,2.5476095199584963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.2,0.5813632011413574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.2,3.09289608001709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.2,1.3248703956604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.2,4.185452651977539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.2,1.5837311744689941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.2,5.286067199707031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.2,0.9875712394714355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.2,6.384921646118164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.2,1.1409791946411132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,1,power_law_1.2,9.609728240966797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.2,1.4990655899047851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.2,1.8197824478149414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,1,power_law_1.2,18.21739501953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.2,2.201318359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,1,power_law_1.2,3.369638442993164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,1,power_law_1.2,5.981139373779297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,2,balanced,0.044026667873064675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,2,balanced,0.044639999667803444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,2,balanced,0.06330133477846782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,2,balanced,0.0919040044148763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,1,power_law_1.2,1.4425087928771974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,2,balanced,0.14727999766667685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,2,balanced,0.1452906628449758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,1,power_law_1.2,1.7088960647583007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,2,balanced,0.13583466410636902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,2,balanced,0.14446933070818582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,1,power_law_1.2,1.9840768814086913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,2,balanced,0.14389866590499878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,1,power_law_1.2,2.5541055679321287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,2,balanced,0.1450399955113729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,2,balanced,0.14043200016021729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,1,power_law_1.2,3.095078468322754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,2,balanced,0.15016532937685648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,1,power_law_1.2,4.232870483398438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,2,balanced,0.14892799655596414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,2,balanced,0.04589866598447164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,2,balanced,0.14589866995811462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,2,balanced,0.05095999936262766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,1,power_law_1.2,5.311059188842774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,2,balanced,0.05966933568318685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,2,balanced,0.1600106656551361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,2,balanced,0.08526399731636047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,1,power_law_1.2,6.3979137420654295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,2,balanced,0.14854933818181357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,2,balanced,0.12693867087364197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,2,balanced,0.04048000027736028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,2,balanced,0.1120746632417043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,2,balanced,0.1527679959932963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,1,power_law_1.2,9.68054428100586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,2,balanced,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,2,balanced,0.11146666606267293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,2,balanced,0.051130667328834534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,2,balanced,0.16064000129699707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,2,balanced,0.10916800300280254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,2,balanced,0.06914666791756947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,2,balanced,0.10467732946077983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,1,power_law_1.2,18.547975158691408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,2,balanced,0.16960533459981283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,2,balanced,0.10059199730555217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,2,balanced,0.1104906698067983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,2,balanced,0.09494933485984802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,2,balanced,0.10872000455856323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,2,balanced,0.09411733349164327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,2,balanced,0.18307199080785116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,2,balanced,0.11263466874758403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,2,balanced,0.0997866690158844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,2,balanced,0.11221866806348164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,2,balanced,0.0963253378868103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,2,balanced,0.11178132891654968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,2,balanced,0.19849065939585367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,2,balanced,0.09844266374905904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,2,balanced,0.12395733594894409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,2,balanced,0.09967466195424397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,2,balanced,0.12441066900889079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,2,balanced,0.09739733735720317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,2,balanced,0.2659253279368083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,2,balanced,0.12214932839075725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,2,balanced,0.10486933588981628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,2,balanced,0.13547733426094055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,2,balanced,0.10123733679453532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,2,balanced,0.11005333065986633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,2,balanced,0.1418613294760386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,2,balanced,0.11037333806355794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,2,balanced,0.2877653241157532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,2,balanced,0.16260799765586853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,2,balanced,0.11432533462842305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,2,balanced,0.124399999777476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,2,balanced,0.1832853356997172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,2,balanced,0.13512000441551208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,2,balanced,0.38510934511820477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,2,balanced,0.264629324277242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,2,balanced,0.1600266695022583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,2,balanced,0.1780853271484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,2,balanced,0.31004265944163006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,2,balanced,0.4861226479212443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,2,balanced,0.24636266628901163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,2,balanced,0.4438186486562093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,2,balanced,0.290501336256663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,2,balanced,0.5745226542154948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,2,balanced,0.6956106821695963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,2,balanced,0.4089653491973877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,2,balanced,0.5289066632588705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,2,balanced,0.8187519709269205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,2,balanced,0.9123413562774658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,2,balanced,0.76583464940389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,2,balanced,1.0373653570810955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,2,balanced,1.0048960049947102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,2,balanced,1.129151980082194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,2,balanced,1.2692426840464275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,2,balanced,1.2426613171895344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,2,balanced,1.9472692807515461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,2,balanced,1.7418559392293294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,2,balanced,0.019920000185569126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,2,balanced,0.022463999688625336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,2,balanced,0.026074667771657307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,2,balanced,0.03124266614516576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,2,balanced,0.04884799818197886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,2,balanced,1.9219573338826497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,2,balanced,0.050144001841545105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,2,balanced,0.05100266635417938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,2,balanced,0.05107733110586802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,2,balanced,0.054325332244237266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,2,balanced,0.054858664671579994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,2,balanced,0.055013333757718406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,2,balanced,0.060880000392595925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,2,balanced,0.062021334966023765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,2,balanced,0.06140799820423126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,2,balanced,0.0810346653064092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,2,balanced,0.07892266909281413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,2,balanced,0.07662400106589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,2,balanced,0.10938133796056111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,2,balanced,0.11834133664766948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,2,balanced,0.13593066732088724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,2,balanced,0.14417066176732382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,2,balanced,0.13885333140691122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,2,balanced,3.770282745361328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,2,balanced,0.1389173368612925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,2,balanced,3.438645362854004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,2,balanced,0.2497333288192749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,2,balanced,0.2395253380139669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,2,balanced,3.719418525695801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,2,balanced,0.3425439993540446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,2,balanced,0.4447253147761027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,2,balanced,0.5474186738332113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,2,balanced,0.8607359727223715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,2,power_law_1.01,0.04382719993591309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,2,power_law_1.01,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,2,power_law_1.01,0.06001920104026794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,2,power_law_1.01,0.07244160175323486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,2,power_law_1.01,0.09064319729804993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,2,power_law_1.01,0.12625919580459594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,2,power_law_1.01,0.13140480518341063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,2,power_law_1.01,0.13493759632110597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,2,power_law_1.01,0.03984000086784363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,2,power_law_1.01,0.13730560541152953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,2,power_law_1.01,0.040582400560379026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,2,power_law_1.01,0.13537280559539794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,2,power_law_1.01,0.04762240052223206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,2,power_law_1.01,0.14385279417037963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,2,power_law_1.01,0.06093440055847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,2,power_law_1.01,0.14364160299301149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,2,power_law_1.01,0.06561279892921448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,2,power_law_1.01,0.1415743947029114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,2,power_law_1.01,0.08626559972763062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,2,power_law_1.01,0.14435839653015137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,2,power_law_1.01,0.08670719861984252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,2,power_law_1.01,0.14910080432891845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,2,power_law_1.01,0.09093120098114013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,2,power_law_1.01,0.15205119848251342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,2,power_law_1.01,0.09176959991455078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,2,power_law_1.01,0.09488000273704529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,2,power_law_1.01,0.15852799415588378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,2,power_law_1.01,0.09605119824409485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,2,power_law_1.01,0.1715648055076599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,2,power_law_1.01,0.09755520224571228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,2,power_law_1.01,0.18851200342178345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,2,balanced,1.6822826067606609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,2,power_law_1.01,0.09918720126152039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,2,power_law_1.01,0.21854720115661622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,2,power_law_1.01,0.10163840055465698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,2,power_law_1.01,0.24350719451904296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,2,power_law_1.01,0.05196160078048706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,2,power_law_1.01,0.1086400032043457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,2,power_law_1.01,0.31308159828186033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,2,power_law_1.01,0.04791040122509003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,2,power_law_1.01,0.10985599756240845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,2,power_law_1.01,0.3646784067153931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,2,power_law_1.01,0.056569600105285646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,2,power_law_1.01,0.11766400337219238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,2,power_law_1.01,0.5245503902435302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,2,power_law_1.01,0.06889600157737732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,2,power_law_1.01,0.13100800514221192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,2,power_law_1.01,0.6032383918762207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,2,power_law_1.01,0.07580159902572632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,2,power_law_1.01,0.14444799423217775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,2,power_law_1.01,0.8400128364562989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,2,power_law_1.01,0.09985920190811157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,2,power_law_1.01,0.17682559490203859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,2,power_law_1.01,1.0438783645629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,2,power_law_1.01,0.10354559421539307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,2,power_law_1.01,0.20921599864959717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,2,power_law_1.01,1.331827163696289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,2,power_law_1.01,0.1046720027923584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,2,power_law_1.01,0.2737152099609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,2,power_law_1.01,0.1072767972946167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,2,power_law_1.01,1.9710975646972657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,2,power_law_1.01,0.32969601154327394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,2,power_law_1.01,0.1074560046195984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,2,power_law_1.01,0.4790976047515869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,2,power_law_1.01,4.170080184936523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,2,power_law_1.01,0.11126400232315063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,2,power_law_1.01,0.5759103775024415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,2,power_law_1.01,0.11245440244674683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,2,power_law_1.01,0.8605183601379395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,2,power_law_1.01,0.11596800088882446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,2,power_law_1.01,1.1370304107666016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,2,power_law_1.01,0.11237119436264038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,2,power_law_1.01,1.4111871719360352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,2,power_law_1.01,0.12149759531021118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,2,power_law_1.01,0.12465920448303222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,2,power_law_1.01,2.1919744491577147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,2,power_law_1.01,0.12819839715957643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,2,power_law_1.01,4.116243362426758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,2,power_law_1.01,0.14939520359039307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,2,power_law_1.01,0.16045440435409547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,2,power_law_1.01,0.1928256034851074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,2,power_law_1.01,0.2265023946762085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,2,power_law_1.01,0.2807424068450928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,2,power_law_1.01,0.3745023965835571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,2,power_law_1.01,0.4907968044281006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,2,power_law_1.01,0.6238592147827149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,2,power_law_1.01,0.949516773223877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,2,power_law_1.01,1.146566390991211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,2,power_law_1.01,1.3466943740844726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,2,power_law_1.01,2.204512023925781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,2,power_law_1.01,4.47248649597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,2,power_law_1.01,0.018534399569034576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,2,power_law_1.01,0.019283199310302736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,2,power_law_1.01,0.023494400084018707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,2,power_law_1.01,0.027584001421928406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,2,power_law_1.01,0.028115200996398925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,2,power_law_1.01,0.04016000032424927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,2,power_law_1.01,0.041945600509643556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,2,power_law_1.01,0.04424319863319397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,2,power_law_1.01,0.045337599515914914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,2,power_law_1.01,0.044972801208496095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,2,power_law_1.2,0.04756479859352112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,2,power_law_1.01,0.04737280011177063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,2,power_law_1.2,0.042931199073791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,2,power_law_1.01,0.049907198548316954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,2,power_law_1.2,0.05067520141601563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,2,power_law_1.01,0.04979200065135956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,2,power_law_1.2,0.0691968023777008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,2,power_law_1.01,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,2,power_law_1.2,0.0819711983203888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,2,power_law_1.01,0.055199998617172244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,2,power_law_1.2,0.1270143985748291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,2,power_law_1.01,0.053388798236846925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,2,power_law_1.2,0.04433279931545257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,2,power_law_1.2,0.13537280559539794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,2,power_law_1.01,0.05816320180892944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,2,power_law_1.2,0.040006399154663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,2,power_law_1.2,0.1408511996269226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,2,power_law_1.01,0.09427199959754944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,2,power_law_1.2,0.04679679870605469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,2,power_law_1.2,0.13789440393447877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,2,power_law_1.2,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,2,power_law_1.2,0.1397055983543396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,2,power_law_1.2,0.06335999965667724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,2,power_law_1.2,0.14202239513397216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,2,power_law_1.2,0.08699520230293274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,2,power_law_1.2,0.14664959907531738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,2,power_law_1.2,0.0899071991443634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,2,power_law_1.2,0.14341119527816773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,2,power_law_1.2,0.09242240190505982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,2,power_law_1.2,0.0939520001411438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,2,power_law_1.2,0.15012480020523072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,2,power_law_1.2,0.09490560293197632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,2,power_law_1.2,0.15687040090560914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,2,power_law_1.2,0.15892479419708253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,2,power_law_1.2,0.09503999948501587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,2,power_law_1.2,0.1643455982208252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,2,power_law_1.2,0.09842560291290284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,2,power_law_1.2,0.1722048044204712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,2,power_law_1.2,0.1
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,2,power_law_1.2,0.1863360047340393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,2,power_law_1.2,0.09939200282096863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,2,power_law_1.2,0.2239232063293457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,2,power_law_1.2,0.11010559797286987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,2,power_law_1.2,0.2540096044540405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,2,power_law_1.2,0.11593600511550903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,2,power_law_1.2,0.12031999826431275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,2,power_law_1.2,0.33359999656677247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,2,power_law_1.2,0.13786879777908326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,2,power_law_1.2,0.37366399765014646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,2,power_law_1.2,0.14953600168228148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,2,power_law_1.2,0.5156799793243408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,2,power_law_1.2,0.18664319515228273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,2,power_law_1.2,0.6523007869720459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,2,power_law_1.2,0.21042559146881104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,2,power_law_1.2,0.8437824249267578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,2,power_law_1.2,0.2924992084503174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,2,power_law_1.2,1.237062358856201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,2,power_law_1.2,0.05144320130348205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,2,power_law_1.2,0.32651519775390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,2,power_law_1.2,1.3872639656066894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,2,power_law_1.2,0.04997119903564453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,2,power_law_1.2,0.5256447792053223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,2,power_law_1.2,0.05724800229072571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,2,power_law_1.2,2.0743616104125975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,2,power_law_1.2,0.6744128227233886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,2,power_law_1.2,0.06727679967880248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,2,power_law_1.2,4.593344116210938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,2,power_law_1.2,0.07382400035858154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,2,power_law_1.2,0.10387840270996093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,2,power_law_1.2,0.10683519840240478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,2,power_law_1.2,0.10792319774627686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,2,power_law_1.2,0.11111680269241334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,2,power_law_1.2,0.11114239692687988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,2,power_law_1.2,0.1123263955116272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,2,power_law_1.2,0.11288959980010986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,2,power_law_1.2,0.11651840209960937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,2,power_law_1.2,0.11731840372085571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,2,power_law_1.2,0.12651519775390624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,2,power_law_1.2,0.1290112018585205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,2,power_law_1.2,0.13212800025939941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,2,power_law_1.2,0.15055359601974488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,2,power_law_1.2,0.1639423966407776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,2,power_law_1.2,0.20468480587005616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,2,power_law_1.2,0.24572160243988037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,2,power_law_1.2,0.29258880615234373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,2,power_law_1.2,0.35916800498962403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,2,power_law_1.2,0.5224256038665771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,2,power_law_1.2,0.5952383995056152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,2,power_law_1.2,0.9741120338439941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,2,power_law_1.2,1.1583807945251465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,2,power_law_1.2,1.457209587097168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,2,power_law_1.2,2.2302528381347657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,2,power_law_1.2,4.951891326904297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,2,power_law_1.01,0.10614399909973145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,2,power_law_1.01,0.10257920026779174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,2,power_law_1.01,0.10241279602050782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,2,power_law_1.01,0.11912959814071655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,2,power_law_1.01,0.13791359663009645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,2,power_law_1.01,0.17375359535217286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,2,power_law_1.01,0.2079296112060547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,2,power_law_1.01,0.2874367952346802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,2,power_law_1.01,0.38674559593200686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,2,power_law_1.2,0.020576000213623047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,2,power_law_1.2,0.020345599949359895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,2,power_law_1.01,0.4453567981719971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,2,power_law_1.2,0.02279680073261261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,2,power_law_1.01,0.6802879810333252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,2,power_law_1.2,0.02643199861049652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,2,power_law_1.01,1.3199040412902832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,2,power_law_1.2,0.028115200996398925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,2,power_law_1.2,0.04222080111503601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,2,power_law_1.2,0.043007999658584595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,2,power_law_1.2,0.04752640128135681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,2,power_law_1.2,0.04542079865932465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,2,power_law_1.2,0.0473471999168396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,2,power_law_1.2,0.04826880097389221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,2,power_law_1.2,0.04702720046043396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,2,power_law_1.2,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,2,power_law_1.2,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,2,power_law_1.2,0.05973119735717773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,2,power_law_1.2,0.058508801460266116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,2,power_law_1.2,0.9508543968200683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,2,power_law_1.2,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,2,power_law_1.2,1.1946047782897948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,2,power_law_1.2,0.09848319888114929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,2,power_law_1.2,1.44203519821167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,2,power_law_1.2,0.1077888011932373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,2,power_law_1.2,2.3076223373413085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,2,power_law_1.2,0.09571840167045594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,2,power_law_1.2,0.09994879961013795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,2,power_law_1.2,4.750060653686523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,2,power_law_1.2,0.12262400388717651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,2,power_law_1.2,0.13925119638442993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,2,power_law_1.2,0.18529280424118041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,2,power_law_1.2,0.22359681129455566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,1,balanced,0.04397333165009817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,2,power_law_1.2,0.3187903881072998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,1,balanced,0.049498667319615684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,2,power_law_1.2,0.4161856174468994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,1,balanced,0.0678719977537791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,1,balanced,0.1002293328444163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,2,power_law_1.2,0.4469759941101074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,1,balanced,0.15427733461062113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,1,balanced,0.14757333199183145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,2,power_law_1.2,0.7295616149902344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,1,balanced,0.14505599935849509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,2,power_law_1.2,1.4643775939941406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,1,balanced,0.14747732877731323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,1,balanced,0.1504693329334259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,1,balanced,0.15292267004648843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,1,balanced,0.1474720040957133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,1,balanced,0.15380799770355225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,1,balanced,0.15522666772206625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,1,balanced,0.15482667088508606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,1,balanced,0.16301866372426352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,1,balanced,0.16359999775886536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,1,balanced,0.16899200280507407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,1,balanced,0.18620800971984863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,1,balanced,0.19737066825230917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,1,balanced,0.22285334269205728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,1,balanced,0.2499306599299113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,1,balanced,0.3380639950434367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,1,balanced,0.38894399007161456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,1,balanced,0.5384000142415365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,1,balanced,0.687994639078776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,1,balanced,0.9967412948608398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,1,balanced,1.31167467435201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,1,balanced,0.0402399996916453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,1,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,1,balanced,0.05359466870625814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,1,balanced,0.07626666625340779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,1,balanced,0.11084266503651936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,1,balanced,0.10854933659235637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,1,balanced,0.10600533088048299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,1,balanced,0.10858133435249329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,1,balanced,0.11051733295122783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,1,balanced,0.10929600397745769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,1,balanced,0.11291733384132385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,1,balanced,0.11395200093587239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,1,balanced,0.1186186671257019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,1,balanced,1.6262772878011067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,1,balanced,0.1199626624584198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,1,balanced,0.1264906624952952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,1,balanced,0.056143999099731445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,1,balanced,0.13292800386746725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,1,balanced,0.06422933439413707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,1,balanced,0.13911466797192892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,1,balanced,0.07540800174077351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,1,balanced,0.10621866583824158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,1,balanced,0.15925332903862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,1,balanced,0.1480959951877594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,1,balanced,0.14275733629862467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,1,balanced,0.1755946675936381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,1,balanced,0.13914666573206583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,1,balanced,0.13470932841300964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,1,balanced,0.13498133420944214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,1,balanced,0.216703991095225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,1,balanced,0.1371999979019165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,1,balanced,0.13914666573206583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,1,balanced,0.24885332584381104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,1,balanced,0.14299733440081278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,1,balanced,0.14290133118629456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,1,balanced,0.14959466457366943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,1,balanced,0.15808533628781637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,1,balanced,0.36133865515391034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,1,balanced,0.16310399770736694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,1,balanced,0.019509332875410717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,1,balanced,0.17006933689117432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,1,balanced,0.020949333906173706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,1,balanced,0.021898667017618816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,1,balanced,0.19620800018310547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,1,balanced,0.025770666698614757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,1,balanced,0.4376000165939331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,1,balanced,0.03249600032965342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,1,balanced,2.518671989440918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,1,balanced,0.21668267250061035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,1,balanced,0.03289066751797994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,1,balanced,0.03438399980465571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,1,balanced,0.2642293373743693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,1,balanced,0.03428266694148382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,1,balanced,0.03483733286460241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,1,balanced,0.03522666543722153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,1,balanced,0.3081386685371399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,1,balanced,0.630570650100708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,1,balanced,0.03835733234882355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,1,balanced,0.0390079990029335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,1,balanced,0.040181333820025124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,1,balanced,0.46540268262227374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,1,balanced,0.04218133290608724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,1,balanced,0.04490133126576742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,1,balanced,0.04438399771849314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,1,balanced,0.5620959997177124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,1,balanced,0.04695466657479604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,1,balanced,0.8223199844360352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,1,balanced,0.05418666700522105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,1,balanced,0.05870933334032694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,1,balanced,0.8160160382588705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,1,balanced,0.0941973328590393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,1,balanced,0.10563733180363973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,1,balanced,1.062442700068156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,1,balanced,0.10363200306892395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,1,balanced,1.1915626525878906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,1,balanced,0.10988266269365947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,1,balanced,1.5421546300252278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,1,balanced,0.18094400564829508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,1,balanced,0.18702399730682373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,1,balanced,1.571781317392985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,1,balanced,2.004074732462565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,1,balanced,4.978085199991862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,1,balanced,0.2609600027402242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,1,balanced,2.45907195409139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,1,balanced,1.954810619354248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,1,balanced,0.3378346761067708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,1,balanced,0.4124799966812134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,1,balanced,3.8454507191975913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,1,balanced,3.054117202758789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,1,balanced,0.6398293177286783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,1,balanced,7.357583999633789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,1,balanced,1.2463946342468262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,1,balanced,5.980485280354817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,1,power_law_1.01,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,1,power_law_1.01,0.046265599131584165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,1,power_law_1.01,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,1,power_law_1.01,0.0762112021446228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,1,power_law_1.01,0.089055997133255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,1,power_law_1.01,0.12447999715805054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,1,power_law_1.01,0.12936960458755492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,1,power_law_1.01,0.13653119802474975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,1,power_law_1.01,0.13747199773788452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,1,power_law_1.01,0.13741439580917358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,1,power_law_1.01,0.14063359498977662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,1,power_law_1.01,0.14256639480590821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,1,power_law_1.01,0.1471168041229248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,1,power_law_1.01,0.14481279850006104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,1,power_law_1.01,0.15136640071868895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,1,power_law_1.01,0.15704959630966187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,1,power_law_1.01,0.1683583974838257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,1,power_law_1.01,0.1839679956436157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,1,power_law_1.01,0.20110080242156983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,1,power_law_1.01,0.24199039936065675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,1,power_law_1.01,0.2742719888687134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,1,power_law_1.01,0.35678720474243164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,1,power_law_1.01,0.4287424087524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,1,power_law_1.01,0.5834112167358398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,1,power_law_1.01,0.7289087772369385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,1,power_law_1.01,1.0228927612304688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,1,power_law_1.01,1.3142144203186035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,1,power_law_1.01,1.6102720260620118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,1,power_law_1.01,2.5007232666015624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,1,power_law_1.01,4.897382354736328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,1,power_law_1.01,0.037920001149177554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,1,power_law_1.01,0.04128639996051788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,1,power_law_1.01,0.049491199851036075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,1,power_law_1.01,0.06222079992294312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,1,power_law_1.01,0.07047680020332336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,1,power_law_1.01,0.09299200177192687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,1,power_law_1.01,0.09683200120925903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,1,power_law_1.01,0.09961599707603455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,1,power_law_1.01,0.10189440250396728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,1,power_law_1.01,0.10092159509658813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,1,power_law_1.01,0.10485119819641113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,1,power_law_1.01,0.10773119926452637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,1,power_law_1.01,0.11242239475250244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,1,power_law_1.01,0.11467519998550416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,1,power_law_1.01,0.12260479927062988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,1,power_law_1.01,0.017343999445438386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,1,power_law_1.01,0.12618240118026733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,1,power_law_1.01,0.017759999632835387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,1,power_law_1.01,0.1384384036064148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,1,power_law_1.01,0.02083200067281723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,1,power_law_1.01,0.0546176016330719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,1,power_law_1.01,0.1605631947517395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,1,power_law_1.01,0.02499839961528778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,1,power_law_1.01,0.05614079833030701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,1,power_law_1.01,0.18188159465789794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,1,power_law_1.01,0.029120001196861266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,1,power_law_1.01,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,1,power_law_1.01,0.227020788192749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,1,power_law_1.01,0.038464000821113585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,1,power_law_1.01,0.08225280046463013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,1,power_law_1.01,0.2675071954727173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,1,power_law_1.01,0.03993600010871887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,1,power_law_1.01,0.0945792019367218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,1,power_law_1.01,0.3648639917373657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,1,power_law_1.01,0.1257472038269043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,1,power_law_1.01,0.04158720076084137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,1,power_law_1.01,0.12757760286331177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,1,power_law_1.01,0.4592832088470459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,1,power_law_1.01,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,1,power_law_1.01,0.12908159494400023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,1,power_law_1.01,0.044863998889923096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,1,power_law_1.01,0.6463232040405273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,1,power_law_1.01,0.13313280344009398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,1,power_law_1.01,0.04663679897785187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,1,power_law_1.01,0.8300800323486328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,1,power_law_1.01,0.1332352042198181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,1,power_law_1.01,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,1,power_law_1.01,1.1893823623657227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,1,power_law_1.01,0.13818880319595336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,1,power_law_1.01,0.04991360008716583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,1,power_law_1.01,1.5486720085144043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,1,power_law_1.01,0.14026880264282227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,1,power_law_1.01,0.05137919783592224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,1,power_law_1.01,1.9193792343139648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,1,power_law_1.01,0.14407680034637452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,1,power_law_1.01,0.05178239941596985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,1,power_law_1.01,0.14601600170135498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,1,power_law_1.01,0.05348479747772217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,1,power_law_1.01,0.15530879497528077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,1,power_law_1.01,0.05556480288505554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,1,power_law_1.01,0.16039040088653564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,1,power_law_1.01,0.061740797758102414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,1,power_law_1.01,0.17342720031738282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,1,power_law_1.01,0.0667136013507843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,1,power_law_1.01,0.2052095890045166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,1,power_law_1.01,0.11793919801712036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,1,power_law_1.01,0.2287424087524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,1,power_law_1.01,0.13187839984893798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,1,power_law_1.01,0.29103999137878417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,1,power_law_1.01,0.1205888032913208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,1,power_law_1.01,0.34761600494384765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,1,power_law_1.01,0.1376960039138794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,1,power_law_1.01,0.47083520889282227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,1,power_law_1.01,0.17775360345840455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,1,power_law_1.01,0.5818240165710449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,1,power_law_1.01,0.2138688087463379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,1,power_law_1.01,0.8170304298400879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,1,power_law_1.01,0.2871232032775879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,1,power_law_1.01,1.0419327735900878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,1,power_law_1.01,0.3667776107788086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,1,power_law_1.01,1.500767993927002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,1,power_law_1.01,0.43809919357299804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,1,power_law_1.01,1.953343963623047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,1,power_law_1.01,0.6698880195617676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,1,power_law_1.01,2.3985664367675783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,1,power_law_1.01,1.271110439300537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,1,power_law_1.01,3.72845458984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,1,power_law_1.01,7.300633239746094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,1,power_law_1.2,0.04158720076084137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,1,power_law_1.2,0.0452672004699707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,1,power_law_1.2,0.05133439898490906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,1,power_law_1.2,0.06963840126991272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,1,power_law_1.2,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,1,power_law_1.2,0.1235967993736267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,1,power_law_1.2,0.13524479866027833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,1,power_law_1.2,0.13822720050811768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,1,power_law_1.2,0.139628803730011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,1,power_law_1.2,0.14004479646682738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,1,power_law_1.2,0.14625920057296754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,1,power_law_1.2,0.14970879554748534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,1,power_law_1.2,0.1524224042892456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,1,power_law_1.2,0.15063040256500243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,1,power_law_1.2,0.15953279733657838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,1,power_law_1.2,0.16470400094985962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,1,power_law_1.2,0.17041280269622802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,1,power_law_1.2,0.1916800022125244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,1,power_law_1.2,0.2093183994293213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,1,power_law_1.2,0.24676480293273925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,1,power_law_1.2,0.27864320278167726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,1,power_law_1.2,0.3584768056869507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,1,power_law_1.2,0.4320256233215332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,1,power_law_1.2,0.5784128189086915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,1,power_law_1.2,0.7233215808868408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,1,power_law_1.2,1.0139264106750487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,1,power_law_1.2,1.3048128128051757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,1,power_law_1.2,1.6030080795288086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,1,power_law_1.2,2.490412712097168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,1,power_law_1.2,4.844473648071289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,1,power_law_1.2,0.05432320237159729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,1,power_law_1.01,2.9997440338134767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,1,power_law_1.2,0.05825279951095581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,1,power_law_1.2,0.06743040084838867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,1,power_law_1.01,5.888371276855469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,1,power_law_1.2,0.07930240035057068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,1,power_law_1.2,0.03839359879493713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,1,power_law_1.2,0.09149439930915833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,1,power_law_1.2,0.04344959855079651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,1,power_law_1.2,0.12306560277938842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,1,power_law_1.2,0.048019200563430786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,1,power_law_1.2,0.1306496024131775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,1,power_law_1.2,0.05767679810523987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,1,power_law_1.2,0.13434239625930786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,1,power_law_1.2,0.13667839765548706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,1,power_law_1.2,0.13768960237503053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,1,power_law_1.2,0.1387712001800537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,1,power_law_1.2,0.14570239782333375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,1,power_law_1.2,0.1469823956489563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,1,power_law_1.2,0.15088000297546386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,1,power_law_1.2,0.1599552035331726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,1,power_law_1.2,0.16517759561538697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,1,power_law_1.2,0.17827839851379396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,1,power_law_1.2,0.20625920295715333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,1,power_law_1.2,0.23233919143676757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,1,power_law_1.2,0.2915328025817871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,1,power_law_1.2,0.34962561130523684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,1,power_law_1.2,0.46494078636169434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,1,power_law_1.2,0.5852223873138428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,1,power_law_1.2,0.8145088195800781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,1,power_law_1.2,1.037945556640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,1,power_law_1.2,1.4916607856750488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,1,power_law_1.2,1.9352575302124024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,1,power_law_1.2,2.3795839309692384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,1,power_law_1.2,3.7146881103515623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,1,power_law_1.2,7.275379180908203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,1,power_law_1.2,0.017478400468826295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,1,power_law_1.2,0.018163199722766876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,1,power_law_1.2,0.02316800057888031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,1,power_law_1.2,0.023827199637889863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,1,power_law_1.2,0.02942720055580139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,1,power_law_1.2,0.03834879994392395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,1,power_law_1.2,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,1,power_law_1.2,0.04298880100250244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,1,power_law_1.2,0.045311999320983884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,1,power_law_1.2,0.04648320078849792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,1,power_law_1.2,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,1,power_law_1.2,0.04811519980430603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,1,power_law_1.2,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,1,power_law_1.2,0.052883201837539674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,1,power_law_1.2,0.05260159969329834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,1,power_law_1.2,0.05418239831924439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,1,power_law_1.2,0.05651199817657471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,1,power_law_1.2,0.06165760159492493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,1,power_law_1.2,0.06796799898147583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,1,power_law_1.2,0.11717760562896729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,1,power_law_1.2,0.13244160413742065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,1,power_law_1.2,0.12372479438781739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,1,power_law_1.2,0.14056960344314576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,1,power_law_1.2,0.182150399684906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,1,power_law_1.2,0.21546878814697265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,1,power_law_1.2,0.29212799072265627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,1,power_law_1.2,0.36981759071350095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,1,power_law_1.2,0.4431295871734619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,1,power_law_1.2,0.6690239906311035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,1,power_law_1.2,1.2763968467712403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,2,balanced,0.06643733382225037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,2,balanced,0.1048906644185384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,2,balanced,0.15787733594576517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,2,balanced,0.26849067211151123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,2,balanced,0.45818666617075604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,2,balanced,0.458682656288147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,2,balanced,0.4469706614812215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,2,balanced,0.4524639844894409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,2,balanced,0.4882346789042155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,2,balanced,0.45861868063608807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,2,balanced,0.4485866626103719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,2,balanced,0.4780319929122925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,2,balanced,0.45552531878153485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,2,balanced,0.452949325243632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,2,balanced,0.4773600101470947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,2,balanced,0.4530880053838094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,2,balanced,0.44651734828948975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,2,balanced,0.4580800135930379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,2,balanced,0.46411200364430744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,2,balanced,0.48657067616780597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,2,balanced,0.5093493461608887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,1,power_law_1.2,0.06848639845848084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,2,balanced,0.7553119659423828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,1,power_law_1.2,0.09624959826469422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,1,power_law_1.2,0.10111360549926758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,1,power_law_1.2,0.10177279710769653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,1,power_law_1.2,0.10414719581604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,2,balanced,0.7291680177052816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,2,balanced,0.05714133381843567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,1,power_law_1.2,0.10382720232009887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,2,balanced,0.07593066493670146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,2,balanced,0.11141866445541382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,1,power_law_1.2,0.10613759756088256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,2,balanced,0.17006933689117432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,1,power_law_1.2,0.11452159881591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,2,balanced,0.2753173311551412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,1,power_law_1.2,0.11519999504089355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,2,balanced,0.2770986755688985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,2,balanced,0.9820160071055094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,1,power_law_1.2,0.11960959434509277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,2,balanced,0.24865599473317465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,1,power_law_1.2,0.12462719678878784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,2,balanced,0.2714879910151164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,1,power_law_1.2,0.1315392017364502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,2,balanced,0.2729919950167338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,2,balanced,0.2651253342628479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,1,power_law_1.2,0.14036480188369752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,2,balanced,0.254533330599467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,1,power_law_1.2,0.16405760049819945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,2,balanced,1.250170628229777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,2,balanced,0.28013867139816284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,1,power_law_1.2,0.18407679796218873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,2,balanced,0.27301865816116333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,1,power_law_1.2,0.2280128002166748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,2,balanced,0.26318933566411334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,1,power_law_1.2,0.27377920150756835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,2,balanced,0.2982026735941569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,1,power_law_1.2,0.3675647974014282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,2,balanced,0.2656426628430684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,1,power_law_1.2,0.4606143951416016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,1,power_law_1.2,0.6436480045318603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,2,balanced,0.27609066168467206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,1,power_law_1.2,0.8237759590148925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,2,balanced,1.831450621287028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,2,balanced,0.28334933519363403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,1,power_law_1.2,1.1879615783691406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,2,balanced,0.2940959930419922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,1,power_law_1.2,1.5438464164733887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,2,balanced,0.3284800052642822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,1,power_law_1.2,1.9056768417358398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,1,power_law_1.2,2.9880319595336915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,2,balanced,0.35920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,1,power_law_1.2,5.883769607543945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,2,balanced,0.5181546608606974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,2,balanced,2.4099626541137695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,2,balanced,0.5410986741383871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,2,balanced,0.748581329981486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,2,balanced,2.9941813151041665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,2,balanced,0.9699680010477701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,2,balanced,1.4003839492797852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,2,balanced,4.785146713256836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,2,balanced,1.8187573750813801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,2,balanced,0.058245331048965454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,2,balanced,0.08388266960779826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,2,balanced,0.12358400225639343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,2,balanced,0.1920693318049113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,2,balanced,0.29531200726826984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,2,balanced,0.2616106669108073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,2,balanced,0.2620159983634949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,2,balanced,0.2693279981613159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,2,balanced,0.26075200239817303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,2,balanced,0.268778661886851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,2,balanced,0.25171732902526855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,2,balanced,0.254528005917867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,2,balanced,0.27797865867614746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,2,balanced,0.25242666403452557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,2,balanced,2.2823840777079263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,2,balanced,0.29370667537053424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,2,balanced,0.26924800872802734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,2,balanced,0.27033599217732746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,2,balanced,0.2752959926923116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,2,balanced,0.29047999779383343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,2,balanced,0.3060533404350281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,2,balanced,0.3298293352127075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,2,balanced,0.5386506716410319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,2,balanced,0.5699413220087687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,2,balanced,3.5838292439778647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,2,balanced,9.604106903076172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,2,balanced,0.7845386664072672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,2,balanced,0.9783573150634766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,2,balanced,1.3809226353963215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,2,balanced,0.02957333376010259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,2,balanced,0.038719999293486275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,2,balanced,0.0525546669960022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,2,balanced,0.07397333284219106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,2,balanced,0.1172106663386027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,2,balanced,0.12032533685366313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,2,balanced,0.12549866239229837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,2,balanced,1.835429350535075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,2,balanced,7.076351801554362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,2,balanced,0.12994133432706198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,2,balanced,0.12962133685747781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,2,balanced,0.13035733501116434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,2,balanced,0.1337493360042572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,2,balanced,0.1644213298956553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,2,balanced,0.14814933141072592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,2,balanced,0.15526400009791055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,2,balanced,0.2456480065981547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,2,balanced,0.2350613276163737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,2,balanced,0.22285866737365723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,2,balanced,0.30982400973637897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,2,balanced,0.33769067128499347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,2,balanced,0.3435200055440267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,2,balanced,2.2531840006510415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,2,balanced,0.34829334417978924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,2,balanced,0.3303413391113281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,2,balanced,0.3276533285776774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,2,balanced,0.5994453430175781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,2,balanced,0.5707093477249146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,2,balanced,3.5080318450927734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,2,balanced,0.8255519866943359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,2,balanced,1.0845653216044109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,2,power_law_1.01,0.10751359462738037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,2,power_law_1.01,0.089055997133255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,2,power_law_1.01,0.12421120405197143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,2,power_law_1.01,0.18773119449615477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,2,power_law_1.01,0.2543231964111328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,2,power_law_1.01,0.3958656072616577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,2,power_law_1.01,0.39491839408874513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,2,power_law_1.01,0.4411327838897705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,2,power_law_1.01,0.41834239959716796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,2,power_law_1.01,0.4408383846282959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,2,power_law_1.01,0.432147216796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,2,balanced,1.3430347442626953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,2,power_law_1.01,0.4408256053924561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,2,power_law_1.01,0.45011200904846194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,2,power_law_1.01,0.4465663909912109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,2,power_law_1.01,0.4509632110595703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,2,power_law_1.01,0.455072021484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,2,power_law_1.01,0.48792319297790526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,2,power_law_1.01,0.5142208099365234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,2,power_law_1.01,0.5730048179626465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,2,power_law_1.01,0.6350912094116211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,2,power_law_1.01,0.7222655773162842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,2,power_law_1.01,0.8856063842773437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,2,power_law_1.01,1.133580780029297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,2,power_law_1.01,1.41910400390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,2,power_law_1.01,1.67509765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,2,power_law_1.01,2.403353691101074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,2,power_law_1.01,2.8933887481689453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,2,balanced,6.877861022949219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,2,power_law_1.01,3.7046016693115233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,2,power_law_1.01,5.724966430664063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,2,balanced,2.122005303700765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,2,power_law_1.01,10.594054412841796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,2,power_law_1.01,0.07662720084190369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,2,power_law_1.01,0.06559360027313232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,2,power_law_1.01,0.08579199910163879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,2,power_law_1.01,0.1298367977142334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,2,power_law_1.01,0.15533440113067626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,2,power_law_1.01,0.2395263910293579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,2,power_law_1.01,0.23795199394226074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,2,power_law_1.01,0.24645121097564698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,2,power_law_1.01,0.2598464012145996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,2,power_law_1.01,0.25553920269012453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,2,power_law_1.01,0.25662078857421877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,2,power_law_1.01,0.2568192005157471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,2,power_law_1.01,0.2686208009719849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,2,power_law_1.01,0.2751935958862305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,2,power_law_1.01,0.28071041107177735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,2,power_law_1.01,0.2765376091003418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,2,power_law_1.01,0.29448959827423093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,2,power_law_1.01,0.31997439861297605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,2,power_law_1.01,0.3465984106063843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,2,power_law_1.01,0.4092735767364502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,2,power_law_1.01,0.49010558128356935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,2,power_law_1.01,0.6100160121917725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,2,power_law_1.01,0.7423488140106201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,2,power_law_1.01,0.9903552055358886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,2,power_law_1.01,1.2021696090698242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,2,power_law_1.01,1.7049407958984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,2,power_law_1.01,2.221638488769531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,2,power_law_1.01,2.8782976150512694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,2,power_law_1.01,4.043622589111328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,2,power_law_1.01,8.4202880859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,2,balanced,4.1965227127075195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,2,power_law_1.01,0.07578240036964416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,2,power_law_1.01,0.026118400692939758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,2,power_law_1.01,0.07336320281028748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,2,power_law_1.01,0.028255999088287354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,2,power_law_1.01,0.04485119879245758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,2,power_law_1.01,0.10532480478286743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,2,power_law_1.01,0.0543936014175415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,2,power_law_1.01,0.13411200046539307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,2,power_law_1.01,0.07049599885940552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,2,power_law_1.01,0.09864320158958435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,2,power_law_1.01,0.10042879581451417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,2,power_law_1.01,0.10623359680175781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,2,power_law_1.01,0.10770560503005981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,2,power_law_1.01,0.10920959711074829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,2,power_law_1.01,0.114956796169281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,2,power_law_1.01,0.11459840536117553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,2,power_law_1.01,0.11631360054016113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,2,power_law_1.01,0.12016639709472657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,2,power_law_1.01,0.14168959856033325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,2,power_law_1.01,0.1411712050437927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,2,power_law_1.01,0.14179840087890624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,2,power_law_1.01,0.2699584007263184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.01,0.30865280628204345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.01,0.22266240119934083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.01,0.24725120067596434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.01,0.2735167980194092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.01,0.31738240718841554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.01,0.4056896209716797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.01,0.5159167766571044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.01,0.6928127765655517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.01,0.8529151916503906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,2,power_law_1.01,1.0908543586730957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,2,power_law_1.2,0.07505919933319091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,2,power_law_1.01,1.6579391479492187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,2,power_law_1.2,0.06360960006713867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,2,power_law_1.01,3.3823680877685547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,2,power_law_1.2,0.08665599822998046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,2,power_law_1.2,0.09483519792556763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,2,power_law_1.2,0.11329920291900634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,2,power_law_1.2,0.07536640167236328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,2,power_law_1.2,0.1466879963874817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,2,power_law_1.2,0.11016960144042968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,2,power_law_1.2,0.2294719934463501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,2,power_law_1.2,0.17663359642028809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,2,power_law_1.2,0.2524480104446411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,2,power_law_1.2,0.2539007902145386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,2,power_law_1.2,0.24666240215301513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,2,power_law_1.2,0.41802239418029785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,2,power_law_1.2,0.2548095941543579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,2,power_law_1.2,0.4344064235687256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,2,power_law_1.2,0.25151360034942627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,2,power_law_1.2,0.4557375907897949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,2,power_law_1.2,0.27086079120635986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,2,power_law_1.2,0.45055999755859377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,2,power_law_1.2,0.27650558948516846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,2,power_law_1.2,0.4413760185241699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,2,power_law_1.2,0.27434239387512205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,2,power_law_1.2,0.26978559494018556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,2,power_law_1.2,0.4618239879608154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,2,power_law_1.2,0.28146560192108155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,2,power_law_1.2,0.46515841484069825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,2,power_law_1.2,0.2946752071380615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,2,power_law_1.2,0.4682112216949463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,2,power_law_1.2,0.29745280742645264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,2,power_law_1.2,0.4740608215332031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,2,power_law_1.2,0.3322751998901367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,2,power_law_1.2,0.49785599708557127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,2,power_law_1.2,0.3632128000259399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,2,power_law_1.2,0.4861248016357422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,2,power_law_1.2,0.43773441314697265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,2,power_law_1.2,0.5074816226959229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,2,power_law_1.2,0.48851838111877444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,2,power_law_1.2,0.5632448196411133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,2,power_law_1.2,0.6474112033843994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,2,power_law_1.2,0.5880320072174072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,2,power_law_1.2,0.739680004119873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,2,power_law_1.2,0.6772096157073975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,2,power_law_1.2,0.7743360042572022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,2,power_law_1.2,1.0290111541748046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,2,power_law_1.2,0.9625344276428223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,2,power_law_1.2,1.2180159568786622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,2,power_law_1.2,1.0769023895263672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,2,power_law_1.2,1.6768320083618165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,2,power_law_1.2,1.4819968223571778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,2,power_law_1.2,2.380563163757324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,2,power_law_1.2,1.7996095657348632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,2,power_law_1.2,2.710495948791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,2,power_law_1.2,2.565132713317871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,2,power_law_1.2,4.143398284912109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,2,power_law_1.2,3.3065025329589846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,2,power_law_1.2,9.432236480712891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,2,power_law_1.2,4.166048049926758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,2,power_law_1.2,6.498239898681641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,2,power_law_1.2,12.758303833007812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,2,power_law_1.01,0.17015039920806885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,2,power_law_1.01,0.23311998844146728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,2,power_law_1.01,0.24330239295959472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,2,power_law_1.01,0.2472831964492798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,2,power_law_1.01,0.24757759571075438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,2,power_law_1.01,0.25706241130828855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,2,power_law_1.01,0.26121599674224855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,2,power_law_1.01,0.2585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,2,power_law_1.01,0.2657088041305542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,2,power_law_1.01,0.2713855981826782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,2,power_law_1.01,0.26944000720977784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,2,power_law_1.01,0.278387188911438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,2,power_law_1.01,0.29637761116027833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,2,power_law_1.01,0.32223360538482665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.01,0.35959041118621826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.01,0.42213120460510256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.01,0.4724287986755371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.01,0.6088319778442383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,2,power_law_1.2,0.07824000120162963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.01,0.7250175952911377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,2,power_law_1.2,0.0648256003856659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.01,0.9688639640808105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,2,power_law_1.2,0.09207040071487427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.01,1.2334976196289062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,2,power_law_1.2,0.13299200534820557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.01,1.6318336486816407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,2,power_law_1.2,0.14742399454116822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.01,2.1740032196044923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,2,power_law_1.2,0.2501888036727905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,2,power_law_1.01,2.4619327545166017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,2,power_law_1.2,0.25024640560150146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,2,power_law_1.2,0.25902719497680665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,2,power_law_1.01,3.942156982421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,2,power_law_1.2,0.26112639904022217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,2,power_law_1.2,0.2644927978515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,2,power_law_1.01,7.141548919677734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,2,power_law_1.2,0.2768383979797363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,2,power_law_1.2,0.2788480043411255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,2,power_law_1.2,0.27721600532531737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,2,power_law_1.2,0.272710394859314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,2,power_law_1.2,0.2858175992965698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,2,power_law_1.2,0.3016256093978882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,2,power_law_1.2,0.3079168081283569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,2,power_law_1.2,0.3389695882797241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.2,0.3629312038421631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,1,balanced,0.0741066684325536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.2,0.44394240379333494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,1,balanced,0.1086079974969228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.2,0.4953279972076416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,1,balanced,0.15755200386047363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.2,0.6249216079711915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,1,balanced,0.27399466435114544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,1,balanced,0.46501866976420086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.2,0.7420928001403808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,2,power_law_1.2,0.027084800601005554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,1,balanced,0.45079465707143146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.2,1.0796480178833008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,2,power_law_1.2,0.02625280022621155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,1,balanced,0.46085866292317706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,2,power_law_1.2,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.2,1.155571174621582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,1,balanced,0.43670399983723956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,2,power_law_1.2,0.05306879878044128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,1,balanced,0.4577600161234538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.2,1.7547136306762696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,1,balanced,0.45555734634399414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,2,power_law_1.2,0.061299198865890504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.2,2.123103904724121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,1,balanced,0.4601066509882609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,2,power_law_1.2,0.10072319507598877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,2,power_law_1.2,2.702560043334961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,1,balanced,0.47153600056966144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,2,power_law_1.2,0.10432640314102173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,1,balanced,0.4567466576894124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,2,power_law_1.2,0.10760960578918458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,2,power_law_1.2,4.558240127563477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,1,balanced,0.47203731536865234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,2,power_law_1.2,0.11107200384140015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,1,balanced,0.45453333854675293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,2,power_law_1.2,0.11290240287780762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,2,power_law_1.2,7.718067169189453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,2,power_law_1.2,0.11545599699020385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,1,balanced,0.4582933187484741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,2,power_law_1.2,0.12137600183486938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,1,balanced,0.47224001089731854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,2,power_law_1.2,0.12113280296325683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,1,balanced,0.48312000433603924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,2,power_law_1.2,0.12711679935455322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,2,power_law_1.2,0.13815679550170898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,1,balanced,0.500053326288859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,2,power_law_1.2,0.14124799966812135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,1,balanced,0.5454506476720175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,1,balanced,0.5864959955215454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,1,balanced,0.8387359778086344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,1,balanced,0.8849226633707682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,1,balanced,1.1975306669871013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,1,balanced,1.5478560129801433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,1,balanced,2.2985706329345703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,1,balanced,3.0265013376871743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,1,balanced,0.0561653325955073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,1,balanced,3.8022985458374023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,1,balanced,0.0765173335870107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,1,balanced,0.1034986674785614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,1,balanced,0.16940265893936157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,1,balanced,0.261957327524821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,1,balanced,0.25940799713134766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,1,balanced,0.255568007628123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,1,balanced,0.26921600103378296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,1,balanced,0.2585866649945577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,1,balanced,0.26597867409388226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,1,balanced,0.2526346643765767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,1,balanced,0.2678933342297872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,1,balanced,0.27821866671244305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,1,balanced,0.27054399251937866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,1,balanced,0.28111465771993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,1,balanced,0.2889653245608012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,1,balanced,0.29705599943796795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,1,balanced,0.3261599938074748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,1,balanced,0.3508373498916626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,1,balanced,0.4089706738789876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,1,balanced,5.925925572713216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,1,balanced,0.45573333899180096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,1,balanced,0.6496426661809286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,1,balanced,0.7408693631490072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,1,balanced,0.07487999896208446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,1,balanced,0.0921013355255127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,1,balanced,0.12684800227483115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,1,balanced,0.19793067375818887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,1,balanced,1.0441226959228516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,1,balanced,0.32280532519022626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,1,balanced,0.28506133953730267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,1,balanced,0.28495999177296955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,1,balanced,0.28777599334716797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,1,balanced,0.2846986651420593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,1,balanced,0.29472533861796063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,1,balanced,0.2864266633987427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,1,balanced,1.358181317647298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,1,balanced,0.28301332394282025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,1,balanced,0.29764799276987713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,1,balanced,0.2880213260650635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,1,balanced,0.30348799626032513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,1,balanced,0.3067466616630554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,1,balanced,0.3094453414281209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,1,balanced,0.33242666721343994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,1,balanced,1.9830506642659504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,1,balanced,0.34748268127441406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,1,balanced,0.40361066659291583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,1,balanced,0.44609065850575763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,1,balanced,0.7231360276540121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,2,power_law_1.2,0.14026880264282227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,1,balanced,2.6263252894083657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,2,power_law_1.2,0.28739199638366697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,1,balanced,0.8030453523000082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.2,0.3156800031661987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,1,balanced,12.02459716796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.2,0.23991680145263672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.2,0.2373055934906006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.2,0.2840512037277222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.2,0.3209023952484131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,1,balanced,1.132368008295695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.2,0.4367487907409668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.2,0.5161344051361084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.2,0.7372159957885742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.2,0.8976767539978028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,1,balanced,3.2526880900065103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,1,balanced,1.443679968516032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,2,power_law_1.2,1.1239551544189452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,2,power_law_1.2,1.710041618347168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,2,power_law_1.2,3.5429054260253907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,1,balanced,2.0910986264546714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,1,balanced,2.7008959452311196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,1,balanced,5.099669456481934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,1,balanced,3.3587519327799478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,1,balanced,5.243573188781738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,1,balanced,10.028575897216797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,1,balanced,0.023669332265853882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,1,balanced,0.026021334032217663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,1,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,1,balanced,0.04952000081539154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,1,balanced,0.07330133517583211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,1,balanced,0.07691733539104462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,1,balanced,0.07905599971612294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,1,balanced,0.08355733752250671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,1,balanced,0.08825600147247314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,1,balanced,0.08998933434486389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,1,balanced,0.09888533751169841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,1,balanced,0.09115200241406758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,1,balanced,0.09379733602205913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,1,balanced,0.10206933816274007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,1,balanced,0.09813867012659709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,1,balanced,0.10098133484522502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,1,balanced,0.10892800490061443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,1,power_law_1.01,0.0718720018863678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,1,balanced,0.11748799681663513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,1,power_law_1.01,0.07698559761047363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,1,balanced,0.12425600488980611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,1,power_law_1.01,0.1275712013244629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,1,power_law_1.01,0.185971200466156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,1,balanced,10.131535847981771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,1,balanced,0.22426666816075644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,1,power_law_1.01,0.25919361114501954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,1,power_law_1.01,0.38156800270080565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,1,power_law_1.01,0.4062528133392334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,1,balanced,0.26550400257110596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,1,power_law_1.01,0.4247039794921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,1,power_law_1.01,0.4210752010345459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,1,power_law_1.01,0.4272448062896729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,1,balanced,0.21130132675170898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,1,power_law_1.01,0.4374527931213379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,1,power_law_1.01,0.4461376190185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,1,power_law_1.01,0.44658560752868653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,1,balanced,0.2174453337987264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,1,power_law_1.01,0.45098237991333007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,1,power_law_1.01,0.4606656074523926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,1,power_law_1.01,0.4726848125457764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,1,power_law_1.01,0.4878079891204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,1,power_law_1.01,0.5338367938995361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,1,power_law_1.01,0.5666111946105957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,1,balanced,0.3720586697260539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,1,power_law_1.01,0.6832831859588623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,1,power_law_1.01,0.7615808010101318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,1,power_law_1.01,0.9532416343688965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,1,power_law_1.01,1.1378687858581542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,1,power_law_1.01,1.49617280960083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,1,power_law_1.01,1.8295616149902343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,1,balanced,0.37187735239664715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,1,power_law_1.01,2.533900833129883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,1,power_law_1.01,3.275040054321289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,1,power_law_1.01,3.9569664001464844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,1,power_law_1.01,6.120825576782226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,1,power_law_1.01,11.893094635009765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,1,balanced,0.5313173135121664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,1,balanced,0.6929813226064047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,1,power_law_1.01,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,1,power_law_1.01,0.06272000074386597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,1,power_law_1.01,0.08566399812698364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,1,power_law_1.01,0.11860480308532714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,1,power_law_1.01,0.14499200582504274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,1,balanced,0.8572586377461752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,1,power_law_1.01,0.22177278995513916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,1,power_law_1.01,0.23032960891723633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,1,power_law_1.01,0.24094080924987793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,1,power_law_1.01,0.23795199394226074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,1,power_law_1.01,0.2445823907852173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,1,power_law_1.01,0.2557248115539551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,1,power_law_1.01,0.2549823999404907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,1,power_law_1.01,0.2687999963760376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,1,power_law_1.01,0.26769280433654785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,1,power_law_1.01,0.2695296049118042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,1,power_law_1.01,0.28563199043273924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,1,power_law_1.01,0.302239990234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,1,power_law_1.01,0.3405951976776123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,1,power_law_1.01,0.3797375917434692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,1,power_law_1.01,0.46440958976745605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,1,power_law_1.01,0.5270847797393798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,1,power_law_1.01,0.7006144046783447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,1,power_law_1.01,0.8365951538085937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,1,power_law_1.01,1.1411775588989257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,1,power_law_1.01,1.4482111930847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,1,balanced,1.3634452819824219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,1,power_law_1.01,2.0522176742553713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,1,power_law_1.01,2.6652799606323243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,1,power_law_1.01,3.262112045288086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,1,power_law_1.01,5.096895980834961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,1,power_law_1.01,10.025183868408202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,1,power_law_1.01,0.06784639954566955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,1,power_law_1.01,0.0760320007801056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,1,power_law_1.01,0.10455039739608765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,1,power_law_1.01,0.14302079677581786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,1,power_law_1.01,0.17049599885940553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,1,power_law_1.01,0.25001599788665774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,1,power_law_1.01,0.26813440322875975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,1,power_law_1.01,0.2718591928482056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,1,power_law_1.01,0.26933119297027586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,1,power_law_1.01,0.2689023971557617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,1,power_law_1.01,0.2798271894454956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,1,power_law_1.01,0.28775041103363036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,1,power_law_1.01,0.29127039909362795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,1,power_law_1.01,0.2940671920776367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,1,power_law_1.01,0.30053119659423827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,1,power_law_1.01,0.3104192018508911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,1,power_law_1.01,0.3274816036224365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,1,power_law_1.01,0.36663041114807127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.01,0.4100287914276123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.01,0.49939842224121095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.01,0.5684800148010254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.01,0.743398380279541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.01,0.8999808311462403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.01,1.2126976013183595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.01,1.5255295753479003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.01,2.1447744369506836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.01,2.7547136306762696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.01,3.4056320190429688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,1,power_law_1.01,5.230368041992188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,1,power_law_1.01,0.021568000316619873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,1,power_law_1.01,0.024454399943351746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,1,power_law_1.01,10.22473602294922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,1,power_law_1.01,0.03438720107078552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,1,balanced,2.6813332239786782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,1,power_law_1.01,0.04671359956264496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,1,power_law_1.01,0.06507520079612732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,1,power_law_1.01,0.09022079706192017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,1,power_law_1.01,0.09500160217285156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,1,power_law_1.01,0.10272639989852905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,1,power_law_1.01,0.10359679460525513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,1,power_law_1.01,0.10812159776687622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,1,power_law_1.01,0.11611520051956177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,1,power_law_1.01,0.10510079860687256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,1,power_law_1.01,0.11212799549102784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,1,power_law_1.01,0.11842559576034546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,1,power_law_1.01,0.11712640523910522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,1,power_law_1.01,0.12318079471588135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,1,power_law_1.01,0.13240959644317626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,1,power_law_1.01,0.13745919466018677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.01,0.15177600383758544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.01,0.3097471952438354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.01,0.3658560037612915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.01,0.250598406791687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.01,0.2808448076248169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.01,0.36752638816833494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.01,0.43427839279174807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.01,0.5945663928985596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.01,0.7498176097869873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.01,0.9080896377563477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,1,power_law_1.01,1.4412416458129882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,1,power_law_1.01,2.7190847396850586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,1,power_law_1.2,0.06780160069465638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,1,power_law_1.2,0.07379199862480164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,1,power_law_1.2,0.12390400171279907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,1,power_law_1.2,0.16673920154571534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,1,power_law_1.2,0.2167680025100708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,1,power_law_1.2,0.3737215995788574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,1,power_law_1.2,0.3911423921585083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,1,power_law_1.2,0.416044807434082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,1,power_law_1.2,0.4278079986572266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,1,power_law_1.2,0.42792320251464844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,1,power_law_1.2,0.434662389755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,1,power_law_1.2,0.44010238647460936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,1,power_law_1.2,0.44701437950134276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,1,power_law_1.2,0.44652161598205564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,1,power_law_1.2,0.4622335910797119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,1,power_law_1.2,0.47688961029052734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,1,power_law_1.2,0.48604159355163573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,1,power_law_1.2,0.5372799873352051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,1,power_law_1.2,0.5820352077484131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,1,power_law_1.2,0.6770559787750244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,1,power_law_1.2,0.7585919857025146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,1,power_law_1.2,0.9408896446228028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,1,power_law_1.2,1.1094464302062987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,1,power_law_1.2,1.4611007690429687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,1,power_law_1.2,1.7987136840820312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,1,power_law_1.2,2.5021184921264648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,1,power_law_1.2,3.195449638366699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,1,power_law_1.2,3.9103103637695313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,1,power_law_1.2,6.027328109741211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,1,power_law_1.2,11.828447723388672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,1,power_law_1.2,0.05331839919090271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,1,power_law_1.2,0.06812800168991089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,1,power_law_1.2,0.06092159748077393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,1,power_law_1.2,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,1,power_law_1.2,0.08269439935684204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,1,power_law_1.2,0.11033600568771362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,1,power_law_1.2,0.10468480587005616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,1,power_law_1.2,0.13666559457778932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,1,power_law_1.2,0.13370879888534545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,1,power_law_1.2,0.16442879438400268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,1,power_law_1.2,0.22851200103759767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,1,power_law_1.2,0.25231359004974363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,1,power_law_1.2,0.24688000679016114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,1,power_law_1.2,0.2756864070892334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,1,power_law_1.2,0.25214080810546874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,1,power_law_1.2,0.285152006149292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,1,power_law_1.2,0.24620161056518555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,1,power_law_1.2,0.283788800239563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,1,power_law_1.2,0.26117119789123533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,1,power_law_1.2,0.2796416044235229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,1,power_law_1.2,0.2603391885757446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,1,power_law_1.2,0.29443840980529784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,1,power_law_1.2,0.2654848098754883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,1,power_law_1.2,0.2926464080810547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,1,power_law_1.2,0.2672640085220337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,1,power_law_1.2,0.26933119297027586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,1,power_law_1.2,0.2960063934326172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,1,power_law_1.2,0.28425600528717043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,1,power_law_1.2,0.3023936033248901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,1,power_law_1.2,0.2943487882614136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,1,power_law_1.2,0.30510079860687256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,1,power_law_1.2,0.30986239910125735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,1,power_law_1.2,0.3195199966430664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,1,power_law_1.2,0.3579967975616455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,1,power_law_1.2,0.33230080604553225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,1,power_law_1.2,0.3952831983566284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,1,power_law_1.2,0.38236799240112307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,1,power_law_1.2,0.4724287986755371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.2,0.4186431884765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,1,power_law_1.2,0.540780782699585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.2,0.4997056007385254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,1,power_law_1.2,0.7020607948303222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.2,0.5748672008514404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,1,power_law_1.2,0.8404992103576661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.2,0.734438419342041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,1,power_law_1.2,1.1547840118408204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.2,0.8978176116943359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,1,power_law_1.2,1.4431296348571778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.2,1.2066880226135255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,1,power_law_1.2,2.0564416885375976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.2,1.5150208473205566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.2,2.1286848068237303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,1,power_law_1.2,2.6523391723632814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.2,2.7290111541748048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,1,power_law_1.2,3.245721435546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.2,3.361094284057617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,1,power_law_1.2,5.087270355224609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,1,power_law_1.2,5.229452896118164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,1,power_law_1.2,10.032089233398438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,1,power_law_1.2,10.190035247802735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,1,power_law_1.2,0.02197760045528412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,1,power_law_1.2,0.027039998769760133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,1,power_law_1.2,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,1,power_law_1.2,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,1,power_law_1.2,0.05990399718284607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,1,power_law_1.2,0.09028480052947999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,1,power_law_1.2,0.10161919593811035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,1,power_law_1.2,0.10067839622497558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,1,power_law_1.2,0.10474879741668701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,1,power_law_1.2,0.10773119926452637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,1,power_law_1.2,0.11825920343399048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,1,power_law_1.2,0.11164799928665162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,1,power_law_1.2,0.1136896014213562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,1,power_law_1.2,0.1227455973625183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,1,power_law_1.2,0.11844480037689209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,1,power_law_1.2,0.12149120569229126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,1,power_law_1.2,0.13235199451446533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,1,power_law_1.2,0.14246400594711303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.2,0.15825920104980468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.2,0.3196415901184082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.2,0.3726655960083008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.2,0.2530112028121948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.2,0.2926975965499878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.2,0.36969599723815916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.2,0.4493887901306152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.2,0.6077760219573974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.2,0.7594560146331787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.2,0.9179391860961914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,1,power_law_1.2,1.456991958618164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,1,power_law_1.2,2.7230527877807615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,2,balanced,0.1456000010172526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,2,balanced,0.2145973245302836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,2,balanced,0.35206933816274005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,2,balanced,0.5756640036900839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,2,balanced,1.0345866680145264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,2,balanced,1.280448039372762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,2,balanced,1.2737332979838054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,2,balanced,1.278704007466634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,2,balanced,1.3179466724395752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,2,balanced,1.3172000249226887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,2,balanced,1.2652746836344402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,2,balanced,1.3161280155181885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,2,balanced,1.3314026991526287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,2,balanced,1.2938026587168376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,2,balanced,1.3189013004302979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,2,balanced,1.3200159867604573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,2,balanced,1.3204800287882488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,2,balanced,1.3186079661051433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,2,balanced,1.3171466986338298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,2,balanced,1.335408051808675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,2,balanced,1.360106627146403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,2,balanced,2.3266080220540366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,2,balanced,2.111423969268799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,2,balanced,3.1691147486368814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,2,balanced,0.09283199906349182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,2,balanced,0.1313920021057129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,2,balanced,0.20148799816767374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,2,balanced,0.33347201347351074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,2,balanced,0.5777493317921957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,2,balanced,0.7234986623128256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,2,balanced,0.7043733596801758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,2,balanced,4.207242647806804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,2,balanced,0.7024479707082113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,2,balanced,0.73089599609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,2,balanced,0.7201333045959473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,2,balanced,0.10989333192507426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,2,balanced,0.7107040087381998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,2,balanced,0.16009066502253214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,2,balanced,0.7479466597239176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,2,balanced,0.2497333288192749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,2,balanced,0.7227146625518799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,2,balanced,0.4059520165125529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,2,balanced,0.7750879923502604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,2,balanced,0.6471946636835734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,2,balanced,0.725594679514567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,2,balanced,0.7476373513539633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,2,balanced,4.482309341430664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,2,balanced,0.6698880195617676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,2,balanced,0.7482826709747314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,2,balanced,0.6795893510182699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,2,balanced,0.6870986620585123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,2,balanced,0.7344640096028646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,2,balanced,0.669818639755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,2,balanced,0.7513493696848551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,2,balanced,0.6964266300201416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,2,balanced,0.7439626852671305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,2,balanced,0.7446186542510986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,2,balanced,0.7143733501434326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,2,balanced,0.7215360005696615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,2,balanced,0.7692639827728271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,2,balanced,0.7474880218505859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,2,balanced,0.7297226587931315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,2,balanced,6.659098943074544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,2,balanced,0.8120693365732828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,2,balanced,0.7268213431040446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,2,balanced,0.7053706645965576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,2,balanced,0.6994453271230062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,2,balanced,1.3603040377298992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,2,balanced,0.7005759874979655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,2,balanced,1.2740853627522786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,2,balanced,0.7037599881490072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,2,balanced,1.3062132994333904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,2,balanced,0.05328533550103506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,2,balanced,6.933712005615234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,2,balanced,0.06816533207893372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,2,balanced,1.9185919761657715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,2,balanced,0.09550399581591289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,2,balanced,0.1499786674976349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,2,balanced,1.2801706790924072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,2,balanced,0.2563199996948242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,2,balanced,0.31838399171829224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,2,balanced,0.32843200365702313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,2,balanced,0.3431413173675537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,2,balanced,0.3521173397699992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,2,balanced,0.33418134848276776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,2,balanced,1.8616533279418945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,2,balanced,0.3475573460261027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,2,balanced,2.569701353708903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,2,balanced,0.3617333173751831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,2,balanced,0.38678932189941406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,2,balanced,0.3847839832305908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,2,balanced,0.3837173382441203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,2,balanced,0.5566720167795817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,2,balanced,2.448549270629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,2,balanced,0.5341600179672241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,2,balanced,0.8244693279266357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,2,balanced,0.8870986302693685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,2,balanced,2.873744010925293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,2,balanced,0.8004586696624756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,2,balanced,11.36092758178711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,2,balanced,0.8207519849141439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,2,balanced,2.7486133575439453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,2,balanced,0.8444746335347494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,2,balanced,0.826533317565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,2,balanced,4.177232106526692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,2,balanced,4.03545602162679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,2,balanced,1.476426601409912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,2,balanced,1.538826624552409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,2,balanced,4.249168078104655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,2,balanced,4.380240122477214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,2,balanced,2.2714719772338867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,2,balanced,23.418785095214844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,2,balanced,2.9229920705159507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,2,balanced,6.907477060953776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,2,balanced,7.119850794474284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,2,balanced,2.929936091105143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,2,balanced,5.304736137390137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,2,balanced,13.724997202555338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,2,balanced,14.380223592122396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,2,balanced,9.366506576538086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,2,power_law_1.01,0.2143615961074829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,2,power_law_1.01,0.2883968114852905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,2,power_law_1.01,0.2649152040481567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,2,power_law_1.01,0.48275198936462405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,2,power_law_1.01,0.6405824184417724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,2,power_law_1.01,0.9979840278625488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,2,power_law_1.01,1.0772671699523926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,2,power_law_1.01,1.1294079780578614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,2,power_law_1.01,1.1050368309020997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,2,power_law_1.01,1.16310396194458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,2,power_law_1.01,1.1877632141113281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,2,power_law_1.01,1.2218688011169434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,2,power_law_1.01,1.1890624046325684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,2,power_law_1.01,1.260095977783203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,2,power_law_1.01,1.248908805847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,2,power_law_1.01,1.2602815628051758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,2,power_law_1.01,1.271763229370117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,2,power_law_1.01,1.3970879554748534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,2,power_law_1.01,1.4908927917480468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,2,power_law_1.01,1.7303743362426758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,2,power_law_1.01,1.8784767150878907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,2,power_law_1.01,2.238918495178223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,2,power_law_1.01,2.7170495986938477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,2,power_law_1.01,3.573535919189453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,2,power_law_1.01,4.18171501159668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,2,power_law_1.01,5.968672180175782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,2,power_law_1.01,7.304716491699219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,2,power_law_1.01,9.00284194946289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,2,power_law_1.01,13.299635314941407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,2,power_law_1.01,27.467193603515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,2,power_law_1.01,0.13867520093917846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,2,power_law_1.01,0.1680896043777466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,2,power_law_1.01,0.17619199752807618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,2,power_law_1.01,0.16368639469146729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,2,power_law_1.01,0.29818239212036135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,2,power_law_1.01,0.2083712100982666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,2,power_law_1.01,0.3816512107849121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,2,power_law_1.01,0.18606719970703126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,2,power_law_1.01,0.5844223976135254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,2,power_law_1.01,0.3036992073059082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,2,power_law_1.01,0.6185215950012207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,2,power_law_1.01,0.36350719928741454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,2,power_law_1.01,0.6115903854370117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,2,power_law_1.01,0.5869696140289307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,2,power_law_1.01,0.6504447937011719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,2,power_law_1.01,0.611078405380249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,2,power_law_1.01,0.6419328212738037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,2,power_law_1.01,0.681382417678833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,2,power_law_1.01,0.6580351829528809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,2,power_law_1.01,0.6873792171478271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,2,power_law_1.01,0.6364863872528076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,2,power_law_1.01,0.6826623916625977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,2,power_law_1.01,0.7009791851043701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,2,power_law_1.01,0.6829184055328369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,2,power_law_1.01,0.6762495994567871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,2,power_law_1.01,0.6990528106689453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,2,power_law_1.01,0.6834176063537598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,2,power_law_1.01,0.7148223876953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,2,power_law_1.01,0.7012095928192139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,2,power_law_1.01,0.7022784233093262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,2,power_law_1.01,0.7152192115783691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,2,power_law_1.01,0.7661824226379395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,2,power_law_1.01,0.049983999133110045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,2,power_law_1.01,0.7096831798553467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,2,power_law_1.01,0.8250304222106933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,2,power_law_1.01,0.0748799979686737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,2,power_law_1.01,0.7407680034637452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,2,power_law_1.01,0.8646464347839355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,2,power_law_1.01,0.0723904013633728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,2,power_law_1.01,0.8037568092346191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,2,power_law_1.01,0.9931967735290528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,2,power_law_1.01,0.12190719842910766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.01,0.8345919609069824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,2,power_law_1.01,1.0984000205993651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,2,power_law_1.01,0.15468159914016724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.01,0.9313983917236328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,2,power_law_1.01,1.422054386138916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,2,power_law_1.01,0.2381119966506958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.01,1.1368000030517578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,2,power_law_1.01,1.6000127792358398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,2,power_law_1.01,0.24649600982666015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.01,1.2894528388977051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,2,power_law_1.01,2.1049856185913085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,2,power_law_1.01,0.278387188911438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.01,1.5542783737182617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,2,power_law_1.01,2.6840320587158204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,2,power_law_1.01,0.27808640003204343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.01,1.9970687866210937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,2,power_law_1.01,3.504198455810547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,2,power_law_1.01,0.2777343988418579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.01,2.461267280578613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,2,power_law_1.01,4.300915145874024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,2,power_law_1.01,0.2897984027862549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.01,3.4653377532958984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,2,power_law_1.01,0.29222400188446046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,2,power_law_1.01,5.5392192840576175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.01,4.248096084594726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,2,power_law_1.01,0.30380799770355227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,2,power_law_1.01,7.99354248046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,2,power_law_1.01,0.3177088022232056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,2,power_law_1.01,5.3267265319824215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,2,power_law_1.01,0.32064640522003174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,2,power_law_1.01,7.579110717773437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,2,power_law_1.01,16.775328063964842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,2,power_law_1.01,0.35948801040649414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,2,power_law_1.01,0.36186239719390867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,2,power_law_1.01,15.162687683105469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,2,power_law_1.01,0.6862847805023193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.01,0.787660789489746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.01,0.5821824073791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.01,0.6011072158813476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.01,0.6945216178894043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.01,0.7439104080200195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.01,0.958681583404541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.01,1.1637568473815918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.01,1.5421055793762206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.01,1.9916671752929687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,2,power_law_1.01,2.4322303771972655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,2,power_law_1.01,3.4960384368896484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,2,power_law_1.01,7.000908660888672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,2,power_law_1.2,0.21566081047058105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,2,power_law_1.2,0.17030400037765503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,2,power_law_1.2,0.266214394569397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,2,power_law_1.2,0.4609344005584717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,2,power_law_1.2,0.5540800094604492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,2,power_law_1.2,1.0077312469482422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,2,power_law_1.2,1.1134783744812011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,2,power_law_1.2,1.1250623703002929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,2,power_law_1.2,1.1622079849243163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,2,power_law_1.2,1.2225600242614747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,2,power_law_1.2,1.2324607849121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,2,power_law_1.2,1.2668160438537597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,2,power_law_1.2,1.2482943534851074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,2,power_law_1.2,1.2865983963012695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,2,power_law_1.2,1.310860824584961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,2,power_law_1.2,1.3463552474975586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,2,power_law_1.2,1.3598143577575683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,2,power_law_1.2,1.519155216217041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,2,power_law_1.2,1.5669568061828614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,2,power_law_1.2,1.805356788635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,2,power_law_1.2,1.9929855346679688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,2,power_law_1.2,2.401241683959961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,2,power_law_1.2,2.8568256378173826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,2,power_law_1.2,3.6275135040283204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,2,power_law_1.2,4.352787017822266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,2,power_law_1.2,5.951980972290039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,2,power_law_1.2,7.411609649658203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,2,power_law_1.2,9.26192626953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,2,power_law_1.2,14.374432373046876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,2,power_law_1.2,29.39591064453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,2,power_law_1.2,0.13235199451446533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,2,power_law_1.2,0.11329920291900634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,2,power_law_1.2,0.1693120002746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,2,power_law_1.2,0.14531840085983277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,2,power_law_1.2,0.23133440017700196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,2,power_law_1.2,0.12502399682998658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,2,power_law_1.2,0.34113919734954834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,2,power_law_1.2,0.18135679960250856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,2,power_law_1.2,0.5840703964233398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,2,power_law_1.2,0.2626431941986084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,2,power_law_1.2,0.6303999900817872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,2,power_law_1.2,0.3336319923400879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,2,power_law_1.2,0.6399231910705566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,2,power_law_1.2,0.5594560146331787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,2,power_law_1.2,0.6811200141906738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,2,power_law_1.2,0.5958335876464844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,2,power_law_1.2,0.6954815864562989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,2,power_law_1.2,0.6407999992370605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,2,power_law_1.2,0.691596794128418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,2,power_law_1.2,0.6348800182342529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,2,power_law_1.2,0.7177984237670898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,2,power_law_1.2,0.6767871856689454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,2,power_law_1.2,0.7172800064086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,2,power_law_1.2,0.6934847831726074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,2,power_law_1.2,0.7035776138305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,2,power_law_1.2,0.6769472122192383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,2,power_law_1.2,0.7427072048187255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,2,power_law_1.2,0.7047296047210694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,2,power_law_1.2,0.7619135856628418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,2,power_law_1.2,0.6975296020507813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,2,power_law_1.2,0.8058431625366211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,2,power_law_1.2,0.7169087886810303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,2,power_law_1.2,0.8429375648498535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,2,power_law_1.2,0.05085440278053284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,2,power_law_1.2,0.7468287944793701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,2,power_law_1.2,0.908409595489502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,2,power_law_1.2,0.0474368005990982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,2,power_law_1.2,0.7439743995666503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,2,power_law_1.2,1.026476764678955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,2,power_law_1.2,0.07263360023498536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,2,power_law_1.2,0.8153727531433106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,2,power_law_1.2,1.1996095657348633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,2,power_law_1.2,0.10269440412521362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.2,0.8780415534973145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,2,power_law_1.2,1.4448512077331543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,2,power_law_1.2,0.13076479434967042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.2,1.0132415771484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,2,power_law_1.2,0.22666239738464355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,2,power_law_1.2,1.6954048156738282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,2,power_law_1.2,0.2624255895614624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.2,1.1574720382690429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,2,power_law_1.2,2.15100154876709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,2,power_law_1.2,0.25895678997039795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.2,1.3911487579345703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,2,power_law_1.2,2.5592832565307617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,2,power_law_1.2,0.29303679466247556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.2,1.6490623474121093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,2,power_law_1.2,3.5677886962890626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,2,power_law_1.2,0.2841408014297485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.2,2.1521984100341798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,2,power_law_1.2,4.700998306274414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,2,power_law_1.2,0.2979135990142822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.2,2.5991872787475585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,2,power_law_1.2,0.3076927900314331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,2,power_law_1.2,5.4628032684326175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.2,3.4148544311523437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,2,power_law_1.2,0.3172096014022827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,2,power_law_1.2,8.340857696533202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,2,power_law_1.2,0.32863359451293944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.2,4.4277183532714846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,2,power_law_1.2,0.33320960998535154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,2,power_law_1.2,5.622521591186524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,2,power_law_1.2,17.448268127441406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,2,power_law_1.2,0.37677440643310545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,2,power_law_1.2,9.276563262939453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,2,power_law_1.2,0.36540799140930175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,2,power_law_1.2,0.7135424137115478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.2,0.8205760002136231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,2,power_law_1.2,16.09742126464844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.2,0.6028800010681152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.2,0.6098624229431152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.2,0.7028096199035645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.2,0.7635072231292724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.2,1.0411007881164551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.2,1.2486207962036133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.2,1.6431808471679688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.2,1.9621376037597655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,2,power_law_1.2,2.40710391998291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,2,power_law_1.2,3.6393665313720702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,2,power_law_1.2,7.144786834716797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,1,balanced,0.14571733276049295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,1,balanced,0.2121760050455729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,1,balanced,0.3280106584231059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,1,balanced,0.5677333275477091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,1,balanced,1.0315093199412029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,1,balanced,1.2687733173370361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,1,balanced,1.2362240155537922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,1,balanced,1.2511413097381592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,1,balanced,1.2876373132069905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,1,balanced,1.280666669209798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,1,balanced,1.2651093006134033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,1,balanced,1.2751200199127197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,1,balanced,1.27292799949646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,1,balanced,1.2724426587422688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,1,balanced,1.2780799865722656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,1,balanced,1.3042133649190266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,1,balanced,1.3046027024586995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,1,balanced,1.3309173583984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,1,balanced,1.3472693761189778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,1,balanced,1.3874880472819011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,1,balanced,1.443413257598877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,1,balanced,2.3519040743509927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,1,balanced,2.190949281056722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,1,balanced,3.348304112752279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,1,balanced,4.4203306833903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,1,balanced,5.032357215881348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,1,balanced,0.0974720021088918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,1,balanced,0.14146666725476584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,1,balanced,0.214464008808136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,1,balanced,0.34298133850097656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,1,balanced,0.5842613379160563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,1,balanced,0.7065386772155762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,1,balanced,7.497802734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,1,balanced,0.6867146492004395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,1,balanced,0.671989361445109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,1,balanced,0.7005066871643066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,1,balanced,0.1157973309357961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,1,balanced,0.7023359934488932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,1,balanced,0.1558026671409607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,1,balanced,0.6932480335235596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,1,balanced,0.23930132389068604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,1,balanced,0.7337919871012369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,1,balanced,0.3824853499730428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,1,balanced,0.7114613056182861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,1,balanced,0.6118346850077311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,1,balanced,0.7640480200449625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,1,balanced,0.7288426558176676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,1,balanced,0.7105600039164225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,1,balanced,0.7058346271514893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,1,balanced,0.6816480159759521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,1,balanced,0.7513973712921143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,1,balanced,0.6910879611968994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,1,balanced,0.6841279665629069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,1,balanced,0.7290080388387045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,1,balanced,0.6559040149052938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,1,balanced,0.7904319763183594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,1,balanced,8.061045328776041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,1,balanced,0.690602699915568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,1,balanced,0.035599999129772186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,1,balanced,0.7057279745737711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,1,balanced,0.7855199972788492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,1,balanced,0.04638933142026266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,1,balanced,0.7216479778289795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,1,balanced,0.061797335743904114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,1,balanced,0.08868799606959026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,1,balanced,0.6915466785430908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,1,balanced,0.852351983388265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,1,balanced,0.14703999956448874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,1,balanced,0.7144426504770914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,1,balanced,0.1801813244819641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,1,balanced,0.1874506672223409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,1,balanced,0.7386666933695475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,1,balanced,0.9220586617787679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,1,balanced,0.19638399283091226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,1,balanced,0.7280213038126627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,1,balanced,0.20362667242685953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,1,balanced,0.21201600631078085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,1,balanced,0.751045306523641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,1,balanced,0.2290666699409485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,1,balanced,1.4283679326375325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,1,balanced,0.24488532543182373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,1,balanced,0.2160586714744568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,1,balanced,0.8047680060068766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,1,balanced,0.23279466231664023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,1,balanced,0.2489759922027588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,1,balanced,1.4565226236979167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,1,balanced,0.8539679845174154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,1,balanced,0.22925867636998495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,1,balanced,0.247381329536438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,1,balanced,0.2814720074335734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,1,balanced,1.4392479260762532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,1,balanced,12.781930287679037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,1,balanced,2.18556801478068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,1,balanced,0.27664534250895184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,1,balanced,0.5492159922917684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,1,balanced,1.5215786298116047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,1,balanced,0.6158666610717773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,1,balanced,2.8658879597981772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,1,balanced,0.48633599281311035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,1,balanced,2.168704032897949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,1,balanced,0.4787786801656087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,1,balanced,3.5060532887776694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,1,balanced,0.8087360064188639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,1,balanced,2.831973393758138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,1,balanced,0.8676053682963053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,1,balanced,4.878005345662435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,1,balanced,3.49615478515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,1,balanced,1.2641226450602214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,1,balanced,26.20660909016927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,1,balanced,5.550032297770183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,1,balanced,1.6360692977905273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,1,balanced,4.85808531443278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,1,balanced,1.701482613881429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,1,balanced,5.492298762003581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,1,balanced,8.826298395792643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,1,balanced,2.91321595509847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,1,balanced,8.784218470255533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,1,balanced,17.75714111328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,1,balanced,5.35975456237793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,1,power_law_1.01,0.1423424005508423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,1,power_law_1.01,0.19054720401763917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,1,power_law_1.01,0.24994559288024903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,1,power_law_1.01,0.4364927768707275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,1,power_law_1.01,0.5880767822265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,1,power_law_1.01,0.9863231658935547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,1,balanced,17.237194061279297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,1,power_law_1.01,1.0177984237670898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,1,power_law_1.01,1.0868736267089845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,1,power_law_1.01,1.1072383880615235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,1,power_law_1.01,1.136672019958496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,1,power_law_1.01,1.1906240463256836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,1,power_law_1.01,1.1745023727416992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,1,power_law_1.01,1.212979221343994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,1,power_law_1.01,1.2196800231933593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,1,power_law_1.01,1.2276800155639649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,1,power_law_1.01,1.2419072151184083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,1,power_law_1.01,1.29748477935791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,1,power_law_1.01,1.4238207817077637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,1,power_law_1.01,1.5087807655334473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,1,power_law_1.01,1.6917951583862305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,1,power_law_1.01,1.8656448364257812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,1,power_law_1.01,2.2670015335083007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,1,power_law_1.01,2.676857566833496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,1,power_law_1.01,3.4067264556884767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,1,power_law_1.01,4.201747131347656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,1,power_law_1.01,5.770572662353516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,1,power_law_1.01,7.296672058105469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,1,power_law_1.01,8.872505950927735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,1,power_law_1.01,13.64520263671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,1,power_law_1.01,27.231103515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,1,power_law_1.01,0.09918720126152039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,1,power_law_1.01,0.13007999658584596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,1,power_law_1.01,0.15679359436035156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,1,power_law_1.01,0.2741760015487671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,1,power_law_1.01,0.3381376028060913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,1,power_law_1.01,0.5488448143005371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,1,power_law_1.01,0.5779839992523194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,1,power_law_1.01,0.5936639785766602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,1,power_law_1.01,0.6167679786682129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,1,power_law_1.01,0.6425792217254639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,1,power_law_1.01,0.6511168003082275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,1,power_law_1.01,0.10758399963378906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,1,power_law_1.01,0.6619584083557128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,1,power_law_1.01,0.14017280340194702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,1,power_law_1.01,0.6690688133239746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,1,power_law_1.01,0.18880000114440917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,1,power_law_1.01,0.6850944042205811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,1,power_law_1.01,0.30512640476226804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,1,power_law_1.01,0.6814720153808593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,1,power_law_1.01,0.35928959846496583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,1,power_law_1.01,0.7033215999603272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,1,power_law_1.01,0.5522240161895752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,1,power_law_1.01,0.7335807800292968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,1,power_law_1.01,0.6204480171203614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,1,power_law_1.01,0.8142592430114746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,1,power_law_1.01,0.6265535831451416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,1,power_law_1.01,0.9002880096435547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,1,power_law_1.01,0.6178175926208496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,1,power_law_1.01,0.9922623634338379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,1,power_law_1.01,0.642086410522461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,1,power_law_1.01,1.1478976249694823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,1,power_law_1.01,0.6532351970672607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,1,power_law_1.01,1.4119359970092773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,1,power_law_1.01,0.6612927913665771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,1,power_law_1.01,1.692460823059082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,1,power_law_1.01,0.6688767910003662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,1,power_law_1.01,0.032358399033546446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,1,power_law_1.01,2.2096128463745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,1,power_law_1.01,0.6771135807037354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,1,power_law_1.01,0.044435200095176694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,1,power_law_1.01,2.7675455093383787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,1,power_law_1.01,0.7115263938903809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,1,power_law_1.01,0.05301759839057922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,1,power_law_1.01,3.786368179321289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,1,power_law_1.01,0.7056000232696533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,1,power_law_1.01,0.09156479835510253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,1,power_law_1.01,4.818041610717773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,1,power_law_1.01,0.13400959968566895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,1,power_law_1.01,0.7452479839324951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,1,power_law_1.01,0.1967743992805481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,1,power_law_1.01,0.8176768302917481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,1,power_law_1.01,5.908908843994141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,1,power_law_1.01,0.24189438819885253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.01,0.8690112113952637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,1,power_law_1.01,9.020985412597657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,1,power_law_1.01,0.2400063991546631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.01,1.0089599609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,1,power_law_1.01,0.2559999942779541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.01,1.146566390991211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,1,power_law_1.01,0.26282880306243894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,1,power_law_1.01,17.853843688964844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.01,1.4140095710754395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,1,power_law_1.01,0.28305280208587646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.01,1.7018815994262695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,1,power_law_1.01,0.31074559688568115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.01,2.2050048828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,1,power_law_1.01,0.2843071937561035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.01,2.7356224060058594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,1,power_law_1.01,0.3037888050079346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.01,3.796352005004883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,1,power_law_1.01,0.317792010307312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,1,power_law_1.01,0.3111936092376709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.01,4.80305290222168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,1,power_law_1.01,0.314246392250061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.01,5.855775833129883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,1,power_law_1.01,0.34380159378051756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.01,0.37791359424591064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,1,power_law_1.01,8.919757080078124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.01,0.7833792209625244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.01,0.9103551864624023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,1,power_law_1.01,17.16890869140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.01,0.57358717918396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.01,0.6349055767059326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.01,0.7913280010223389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.01,0.934547233581543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.01,1.2640192031860351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.01,1.5887935638427735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.01,1.8777664184570313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,1,power_law_1.01,2.8706111907958984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,1,power_law_1.2,0.1416383981704712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,1,power_law_1.01,5.458425521850586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,1,power_law_1.2,0.15070079565048217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,1,power_law_1.2,0.2325119972229004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,1,power_law_1.2,0.37864320278167723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,1,power_law_1.2,0.5100287914276123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,1,power_law_1.2,0.9739392280578614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,1,power_law_1.2,1.0577024459838866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,1,power_law_1.2,1.0573439598083496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,1,power_law_1.2,1.1327872276306152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,1,power_law_1.2,1.1514623641967774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,1,power_law_1.2,1.1926207542419434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,1,power_law_1.2,1.2113408088684081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,1,power_law_1.2,1.235961627960205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,1,power_law_1.2,1.241260814666748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,1,power_law_1.2,1.2764991760253905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,1,power_law_1.2,1.2879615783691407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,1,power_law_1.2,1.3560064315795899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,1,power_law_1.2,1.456550407409668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,1,power_law_1.2,1.5461119651794433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,1,power_law_1.2,1.738105583190918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,1,power_law_1.2,1.947808074951172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,1,power_law_1.2,2.3270912170410156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,1,power_law_1.2,2.69036808013916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,1,power_law_1.2,3.4648193359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,1,power_law_1.2,4.222860717773438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,1,power_law_1.2,5.705696105957031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,1,power_law_1.2,7.235737609863281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,1,power_law_1.2,8.844831848144532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,1,power_law_1.2,13.545484924316407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,1,power_law_1.2,26.5466552734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,1,power_law_1.2,0.10177279710769653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,1,power_law_1.2,0.10274560451507568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,1,power_law_1.2,0.1564352035522461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,1,power_law_1.2,0.2396928071975708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,1,power_law_1.2,0.3065664052963257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,1,power_law_1.2,0.5461631774902344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,1,power_law_1.2,0.5786240100860596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,1,power_law_1.2,0.5999104022979737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,1,power_law_1.2,0.6402624130249024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,1,power_law_1.2,0.661836814880371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,1,power_law_1.2,0.6690815925598145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,1,power_law_1.2,0.11655679941177369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,1,power_law_1.2,0.6831615924835205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,1,power_law_1.2,0.12002559900283813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,1,power_law_1.2,0.6831488132476806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,1,power_law_1.2,0.1776576042175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,1,power_law_1.2,0.6922368049621582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,1,power_law_1.2,0.26963200569152834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,1,power_law_1.2,0.721017599105835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,1,power_law_1.2,0.3241408109664917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,1,power_law_1.2,0.7289279937744141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,1,power_law_1.2,0.5696640014648438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,1,power_law_1.2,0.7636928081512451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,1,power_law_1.2,0.5869631767272949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,1,power_law_1.2,0.8434047698974609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,1,power_law_1.2,0.6316160202026367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,1,power_law_1.2,0.9095359802246094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,1,power_law_1.2,0.6305535793304443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,1,power_law_1.2,1.048851203918457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,1,power_law_1.2,0.6422912120819092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,1,power_law_1.2,1.159705638885498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,1,power_law_1.2,0.03175680041313171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,1,power_law_1.2,0.683462381362915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,1,power_law_1.2,0.03640320003032684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,1,power_law_1.2,1.4433279991149903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,1,power_law_1.2,0.6934271812438965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,1,power_law_1.2,0.051801598072052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,1,power_law_1.2,1.6990144729614258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,1,power_law_1.2,0.7019904136657715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,1,power_law_1.2,0.08608639836311341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,1,power_law_1.2,2.2216512680053713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,1,power_law_1.2,0.7127423763275147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,1,power_law_1.2,0.12284799814224243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,1,power_law_1.2,2.748371124267578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,1,power_law_1.2,0.20646400451660157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,1,power_law_1.2,0.735200023651123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,1,power_law_1.2,3.780140686035156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,1,power_law_1.2,0.23399040699005128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,1,power_law_1.2,0.7335999965667724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,1,power_law_1.2,0.24109439849853515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,1,power_law_1.2,4.814035034179687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,1,power_law_1.2,0.763430404663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,1,power_law_1.2,0.26026239395141604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,1,power_law_1.2,0.8448320388793945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,1,power_law_1.2,5.84727668762207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,1,power_law_1.2,0.2715712070465088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.2,0.9039360046386719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,1,power_law_1.2,0.2891135931015015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,1,power_law_1.2,9.016044616699219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.2,1.0441216468811034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,1,power_law_1.2,0.319705605506897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.2,1.171609592437744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,1,power_law_1.2,0.2928895950317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.2,1.4309311866760255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,1,power_law_1.2,17.429017639160158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,1,power_law_1.2,0.313427209854126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.2,1.6907583236694337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,1,power_law_1.2,0.33152639865875244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.2,2.2313791275024415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,1,power_law_1.2,0.3082304000854492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.2,2.7159551620483398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,1,power_law_1.2,0.3341504096984863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.2,3.7501056671142576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,1,power_law_1.2,0.3615295886993408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.2,0.38224000930786134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.2,4.788083267211914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.2,0.7974976062774658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.2,5.824319839477539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.2,0.9340736389160156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.2,0.5965760231018067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,1,power_law_1.2,8.839775848388673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.2,0.66561918258667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.2,0.8080448150634766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,1,power_law_1.2,17.01329345703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.2,0.9640255928039551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.2,1.2755135536193847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.2,1.587052822113037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.2,1.9096384048461914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,1,power_law_1.2,2.860767936706543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,1,power_law_1.2,5.47470703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,2,balanced,0.12713066736857095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,2,balanced,0.20610666275024414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,2,balanced,0.3393813371658325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,2,balanced,0.5489706595738729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,2,balanced,0.9826080004374186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,2,balanced,1.8469759623209636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,2,balanced,2.6500372886657715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,2,balanced,2.6559786796569824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,2,balanced,2.6597867012023926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,2,balanced,2.732858657836914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,2,balanced,2.7265707651774087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,2,balanced,2.7058614095052085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,2,balanced,2.671482721964518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,2,balanced,2.7262185414632163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,2,balanced,2.698906580607096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,2,balanced,2.674762725830078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,2,balanced,2.768954594930013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,2,balanced,2.7354933420817056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,2,balanced,2.794053395589193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,2,balanced,2.7727041244506836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,2,balanced,2.8024266560872397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,2,balanced,2.862160046895345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,2,balanced,2.930389404296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,2,balanced,3.108037312825521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,2,balanced,5.0251468022664385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,2,balanced,4.691184043884277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,2,balanced,6.832357406616211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,2,balanced,0.09745599826176961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,2,balanced,0.13733866810798645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,2,balanced,0.20816532770792642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,2,balanced,0.3565280040105184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,2,balanced,0.577616016070048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,2,balanced,0.9886666933695475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,2,balanced,1.4647520383199055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,2,balanced,1.440293312072754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,2,balanced,1.4112693468729656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,2,balanced,1.4003307024637859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,2,balanced,1.4092373847961426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,2,balanced,1.4535519282023113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,2,balanced,1.432586669921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,2,balanced,9.697258631388346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,2,balanced,1.4528106053670247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,2,balanced,1.4505546887715657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,2,balanced,1.4797333081563313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,2,balanced,0.11389866471290588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,2,balanced,1.5173120498657227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,2,balanced,0.16315199931462607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,2,balanced,0.21991999944051108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,2,balanced,1.4618453979492188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,2,balanced,0.37241601943969727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,2,balanced,0.6138079961140951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,2,power_law_1.01,0.20422399044036865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,2,balanced,1.5191253026326497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,2,balanced,1.0182452996571858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,2,power_law_1.01,0.3368832111358643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,2,balanced,1.4350666999816895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,2,balanced,1.4273279507954915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,2,balanced,1.5382186571757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,2,power_law_1.01,0.5761536121368408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,2,balanced,1.384383996327718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,2,power_law_1.01,0.42711682319641114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,2,balanced,1.3382666905721028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,2,balanced,1.585594654083252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,2,power_law_1.01,0.7975872039794922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,2,balanced,1.3301653067270915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,2,balanced,1.3242186705271404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,2,power_law_1.01,1.189030361175537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,2,balanced,1.2967092990875244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,2,power_law_1.01,1.393228816986084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,2,balanced,1.3058773676554363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,2,balanced,1.6608373324076335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,2,power_law_1.01,1.7463680267333985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,2,balanced,1.4212907155354817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,2,balanced,14.041173299153646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,2,power_law_1.01,1.8347583770751954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,2,balanced,1.3956480026245117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,2,power_law_1.01,1.982374382019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,2,balanced,1.4129066467285156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,2,balanced,1.7428213755289714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,2,power_law_1.01,1.9203199386596679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,2,balanced,1.3767840067545574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,2,power_law_1.01,2.039219284057617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,2,balanced,1.3735307057698567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,2,power_law_1.01,2.1125696182250975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,2,power_law_1.01,2.1335935592651367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,2,balanced,1.41703462600708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,2,balanced,1.959546724955241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,2,power_law_1.01,2.1866687774658202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,2,balanced,1.4808905919392903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,2,power_law_1.01,2.234272003173828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,2,power_law_1.01,2.2985984802246096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,2,power_law_1.01,2.397881507873535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,2,balanced,3.03817081451416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,2,balanced,1.4811894098917644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,2,power_law_1.01,2.5084415435791017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,2,power_law_1.01,2.6796735763549804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,2,power_law_1.01,2.952102470397949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,2,balanced,1.528506596883138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,2,power_law_1.01,3.336102294921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,2,power_law_1.01,3.713670349121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,2,power_law_1.01,4.616831970214844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,2,balanced,3.0941012700398765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,2,power_law_1.01,5.3206336975097654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,2,balanced,1.7255999247233074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,2,power_law_1.01,7.042578887939453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,2,power_law_1.01,8.546745300292969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,2,power_law_1.01,9.633695983886719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,2,power_law_1.01,14.688844299316406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,2,balanced,2.9150079091389975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,2,balanced,4.386490821838379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,2,power_law_1.01,27.65244140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,2,balanced,23.796417236328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,2,balanced,2.9493227005004883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,2,balanced,6.0676320393880205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,2,balanced,4.188085238138835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,2,balanced,8.848410924275717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,2,balanced,5.68719482421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,2,balanced,8.41585604349772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,2,balanced,15.691445668538412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,2,power_law_1.01,0.1313472032546997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,2,power_law_1.01,0.2060096025466919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,2,power_law_1.01,0.14576640129089355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,2,power_law_1.01,0.32413439750671386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,2,power_law_1.01,0.24180479049682618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,2,power_law_1.01,0.23439359664916992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,2,power_law_1.01,0.3497920036315918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,2,power_law_1.01,0.4322048187255859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,2,power_law_1.01,0.2653887987136841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,2,power_law_1.01,0.6212800025939942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,2,power_law_1.01,0.5036159992218018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,2,power_law_1.01,0.7476096153259277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,2,power_law_1.01,0.6411263942718506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,2,power_law_1.01,0.9487744331359863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,2,power_law_1.01,0.748633623123169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,2,power_law_1.01,0.9673600196838379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,2,power_law_1.01,0.9026432037353516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,2,power_law_1.01,1.0281344413757325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,2,power_law_1.01,0.9742591857910157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,2,power_law_1.01,1.070867156982422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,2,power_law_1.01,0.992467212677002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,2,power_law_1.01,1.071833610534668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,2,power_law_1.01,1.0418047904968262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,2,power_law_1.01,1.1194815635681152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,2,power_law_1.01,1.1171072006225586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,2,power_law_1.01,1.0887871742248536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,2,power_law_1.01,1.1484607696533202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,2,power_law_1.01,1.194803237915039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,2,power_law_1.01,1.1499327659606933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,2,power_law_1.01,1.1902527809143066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,2,power_law_1.01,1.1566975593566895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,2,power_law_1.01,1.2650495529174806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,2,power_law_1.01,1.1840895652770995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,2,power_law_1.01,1.1960384368896484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,2,power_law_1.01,1.291539192199707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,2,power_law_1.01,1.2300992012023926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,2,power_law_1.01,1.336012840270996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,2,balanced,14.810101826985678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,2,power_law_1.01,1.2933568000793456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,2,power_law_1.01,1.5220735549926758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,2,power_law_1.01,1.4090751647949218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,2,power_law_1.01,1.66046085357666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,2,power_law_1.01,1.5325887680053711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,2,power_law_1.01,1.8795072555541992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,2,power_law_1.01,1.798464012145996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,2,power_law_1.01,2.157644844055176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,2,power_law_1.01,2.6083967208862306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,2,power_law_1.01,2.017862319946289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,2,power_law_1.01,2.5318784713745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,2,power_law_1.01,3.1061376571655273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,2,power_law_1.01,3.0128704071044923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,2,power_law_1.01,4.221875381469727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,2,power_law_1.01,3.885753631591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,2,power_law_1.01,5.167923355102539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,2,power_law_1.01,4.926150512695313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,2,power_law_1.01,6.044441604614258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,2,power_law_1.01,5.691091156005859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,2,power_law_1.01,9.063219451904297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,2,power_law_1.01,8.634598541259766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,2,power_law_1.01,16.813856506347655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,2,power_law_1.01,15.864044189453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,2,power_law_1.2,0.20313599109649658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,2,power_law_1.2,0.3230528116226196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,2,power_law_1.2,0.3683903932571411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,2,power_law_1.2,0.4045504093170166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,2,power_law_1.2,0.7165760040283203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,2,power_law_1.2,1.0063679695129395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,2,power_law_1.2,1.1965632438659668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,2,power_law_1.2,1.6208383560180664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,2,power_law_1.2,1.7409536361694335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,2,power_law_1.2,1.8389503479003906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,2,power_law_1.2,1.9128768920898438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,2,power_law_1.2,1.952012825012207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,2,power_law_1.2,2.084992027282715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,2,power_law_1.2,2.0848703384399414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,2,power_law_1.2,2.132499122619629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,2,power_law_1.2,2.178425598144531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,2,power_law_1.2,2.2493696212768555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,2,power_law_1.2,2.4292224884033202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,2,power_law_1.2,2.581900787353516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,2,power_law_1.2,2.8236799240112305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,2,power_law_1.2,3.10001277923584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,2,power_law_1.2,3.5195327758789063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,2,power_law_1.2,3.978547286987305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,2,power_law_1.2,0.1344063997268677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,2,power_law_1.2,0.20346879959106445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,2,power_law_1.2,4.89148178100586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,2,power_law_1.2,0.22524800300598144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,2,power_law_1.2,5.552115249633789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,2,power_law_1.2,0.26936318874359133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,2,power_law_1.2,7.19815673828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,2,power_law_1.2,0.42256641387939453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,2,power_law_1.2,0.5806272029876709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,2,power_law_1.2,9.00600357055664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,2,power_law_1.2,0.6927616119384765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,2,power_law_1.2,10.552159881591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,2,power_law_1.2,0.9066304206848145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,2,power_law_1.2,0.9877504348754883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,2,power_law_1.2,15.38828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,2,power_law_1.2,0.9521344184875489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,2,power_law_1.2,1.0555071830749512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,2,power_law_1.2,1.0933631896972655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,2,power_law_1.2,29.642105102539062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,2,power_law_1.2,1.1300031661987304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,1,balanced,0.13379733761151633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,2,power_law_1.2,1.156447982788086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,1,balanced,0.20536533991495767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,2,power_law_1.2,0.16419199705123902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,2,power_law_1.2,1.209228801727295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,1,balanced,0.3271413246790568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,1,balanced,0.5366933345794678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,2,power_law_1.2,0.2185983896255493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,2,power_law_1.2,1.2095935821533204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,1,balanced,0.9611732959747314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,2,power_law_1.2,0.24606719017028808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,2,power_law_1.2,1.271615982055664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,1,balanced,1.783690611521403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,1,balanced,2.6041226387023926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,2,power_law_1.2,0.28733439445495607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,2,power_law_1.2,1.3408384323120117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,1,balanced,2.633135954538981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,2,power_law_1.2,0.4308351993560791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,2,power_law_1.2,1.4179200172424316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,1,balanced,2.6184372901916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,2,power_law_1.2,0.5826367855072021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,2,power_law_1.2,1.6013887405395508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,1,balanced,2.623151938120524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,1,balanced,2.6694558461507163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,2,power_law_1.2,0.6868415832519531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,2,power_law_1.2,1.739776039123535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,1,balanced,2.66869322458903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,2,power_law_1.2,0.8682880401611328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,2,power_law_1.2,2.046291160583496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,1,balanced,2.6543307304382324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,2,power_law_1.2,0.945740795135498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,2,power_law_1.2,2.2944320678710937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,1,balanced,2.6612799962361655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,2,power_law_1.2,0.9758591651916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,1,balanced,2.6654186248779297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,2,power_law_1.2,2.872659111022949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,1,balanced,2.7010933558146157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,2,power_law_1.2,1.065120029449463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,2,power_law_1.2,3.3077312469482423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,1,balanced,2.687077204386393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,2,power_law_1.2,1.0697407722473145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,2,power_law_1.2,4.327353668212891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,2,power_law_1.2,1.1155263900756835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,1,balanced,2.7151145935058594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,2,power_law_1.2,5.197407913208008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,2,power_law_1.2,1.1365568161010742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,1,balanced,2.8036321004231772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,2,power_law_1.2,6.663359832763672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,2,power_law_1.2,1.1462976455688476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,1,balanced,2.8160746892293296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,2,power_law_1.2,1.2110464096069335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,2,power_law_1.2,9.679590606689453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,2,power_law_1.2,1.2157440185546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,1,balanced,2.9043893814086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,2,power_law_1.2,1.31461763381958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,2,power_law_1.2,17.986636352539062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,2,power_law_1.2,1.3788928031921386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,2,power_law_1.2,1.51015043258667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,1,balanced,3.013413429260254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,2,power_law_1.2,1.724518394470215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,2,power_law_1.2,1.9169151306152343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,1,balanced,3.1428426106770835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,2,power_law_1.2,2.2712064743041993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,2,power_law_1.2,2.682464027404785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,2,power_law_1.2,3.211558532714844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,2,power_law_1.2,4.2011455535888675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,1,balanced,3.4521919886271157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,2,power_law_1.2,5.018873596191407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,2,power_law_1.2,6.19365119934082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,2,power_law_1.2,8.598387145996094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,2,power_law_1.2,16.901408386230468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,1,balanced,5.3391469319661455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,1,balanced,5.319039980570476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,1,balanced,7.51469357808431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,1,balanced,10.28166389465332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,1,balanced,0.10141866405804952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,1,balanced,0.13499733805656433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,1,balanced,0.21524266401926676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,1,balanced,0.33750398953755695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,1,balanced,0.5686666568120321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,1,balanced,1.0046772956848145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,1,balanced,1.4375252723693848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,1,balanced,1.440442721048991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,1,balanced,1.4183947245279949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,1,balanced,1.4104372660319011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,1,balanced,1.4118506113688152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,1,balanced,1.403509298960368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,1,balanced,1.4256426493326824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,1,balanced,1.4114346504211426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,1,balanced,1.4633760452270508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,1,balanced,1.4693919817606609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,1,balanced,1.483407974243164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,1,balanced,1.5254933039347331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,1,balanced,1.547317345937093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,1,balanced,15.217498779296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,1,balanced,1.6275466283162434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,1,balanced,1.6935787200927734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,1,balanced,1.837941328684489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,1,balanced,0.14127467075983682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,1,balanced,0.1720586617787679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,1,balanced,0.25043733914693195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,1,balanced,0.39427733421325684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,1,balanced,0.6122346719106039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,1,balanced,2.013045310974121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,1,balanced,1.0725226402282715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,1,balanced,1.4833440780639648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,1,balanced,1.4857813517252605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,1,balanced,1.4579946200052898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,1,balanced,1.412314732869466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,1,balanced,2.368410587310791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,1,balanced,1.4108853340148926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,1,balanced,1.3856533368428547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,1,balanced,1.365429401397705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,1,balanced,1.3800907135009766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,1,balanced,1.4289546012878418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,1,balanced,1.4044639269510906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,1,balanced,3.4261598587036133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,1,balanced,1.463029384613037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,1,balanced,1.4512054125467937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,1,balanced,1.498143990834554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,1,power_law_1.01,0.12406400442123414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,1,balanced,1.545680046081543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,1,power_law_1.01,0.19351040124893187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,1,power_law_1.01,0.3129983901977539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,1,balanced,3.899386723836263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,1,balanced,1.6336533228556316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,1,power_law_1.01,0.36727681159973147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,1,power_law_1.01,0.7681024074554443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,1,balanced,1.737552007039388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,1,power_law_1.01,1.0527999877929688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,1,power_law_1.01,1.2570367813110352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,1,balanced,1.8835360209147136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,1,power_law_1.01,1.6951936721801757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,1,power_law_1.01,1.7643264770507812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,1,power_law_1.01,1.805638313293457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,1,balanced,5.358901341756185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,1,power_law_1.01,1.9188224792480468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,1,balanced,2.273866653442383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,1,balanced,27.732874552408855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,1,power_law_1.01,1.9641151428222656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,1,power_law_1.01,2.016377639770508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,1,power_law_1.01,2.102374458312988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,1,power_law_1.01,2.1461183547973635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,1,balanced,3.5290454228719077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,1,power_law_1.01,2.156697654724121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,1,power_law_1.01,2.24322566986084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,1,power_law_1.01,2.3832000732421874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,1,power_law_1.01,2.4495744705200195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,1,balanced,7.0399519602457685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,1,power_law_1.01,2.6831743240356447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,1,power_law_1.01,2.8949951171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,1,balanced,3.971482594807943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,1,power_law_1.01,3.272447967529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,1,power_law_1.01,3.687001419067383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,1,power_law_1.01,4.559404754638672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,1,power_law_1.01,5.402054214477539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,1,power_law_1.01,7.146636962890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,1,power_law_1.01,8.656928253173827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,1,balanced,5.525583902994792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,1,power_law_1.01,10.261548614501953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,1,power_law_1.01,15.205094909667968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,1,power_law_1.01,28.324557495117187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,1,balanced,10.568826675415039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,1,balanced,7.210874557495117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,1,balanced,10.866570790608725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,1,balanced,19.712591807047527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,1,power_law_1.01,0.09287040233612061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,1,power_law_1.01,0.12989439964294433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,1,power_law_1.01,0.20188159942626954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,1,power_law_1.01,0.24508800506591796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,1,power_law_1.01,0.4195456027984619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,1,power_law_1.01,0.598521614074707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,1,power_law_1.01,0.7073984146118164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,1,balanced,19.39356740315755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,1,power_law_1.01,0.897702407836914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,1,power_law_1.01,0.9351743698120117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,1,power_law_1.01,0.9740032196044922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,1,power_law_1.01,1.0427776336669923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,1,power_law_1.01,1.06364164352417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,1,power_law_1.01,1.0784768104553222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,1,power_law_1.01,1.1484864234924317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,1,power_law_1.01,1.1581055641174316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,1,power_law_1.01,1.1993599891662599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,1,power_law_1.01,1.2204671859741212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,1,power_law_1.01,1.3185600280761718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,1,power_law_1.01,1.3878080368041992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,1,power_law_1.01,0.13457920551300048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,1,power_law_1.01,1.535968017578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,1,power_law_1.01,0.17552640438079833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,1,power_law_1.01,1.6672319412231444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,1,power_law_1.01,0.23564798831939698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,1,power_law_1.01,1.9413631439208985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,1,power_law_1.01,0.28910079002380373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,1,power_law_1.01,2.2256832122802734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,1,power_law_1.01,0.46581759452819826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,1,power_law_1.01,2.858284759521484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,1,power_law_1.01,0.6755712032318115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,1,power_law_1.01,3.3944255828857424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,1,power_law_1.01,0.7498623847961425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,1,power_law_1.01,4.534067153930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,1,power_law_1.01,0.9795392036437989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,1,power_law_1.01,1.0126144409179687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,1,power_law_1.01,5.707968139648438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,1,power_law_1.01,1.053382396697998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,1,power_law_1.01,6.855506896972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,1,power_law_1.01,1.0541824340820312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,1,power_law_1.01,10.353081512451173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,1,power_law_1.01,1.112384033203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,1,power_law_1.01,1.1130111694335938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,1,power_law_1.01,1.1498944282531738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,1,power_law_1.01,19.660755920410157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,1,power_law_1.01,1.1979455947875977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,1,power_law_1.01,1.2186240196228026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,1,power_law_1.01,1.2455936431884767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,1,power_law_1.01,1.311398410797119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.01,1.3610431671142578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.01,1.5338815689086913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.01,1.6403200149536132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.01,1.9579200744628906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.01,2.260268783569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.01,2.8620607376098635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,1,power_law_1.01,3.459379196166992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,1,power_law_1.01,4.662201690673828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,1,power_law_1.01,5.8412223815917965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,1,power_law_1.01,6.982463836669922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,1,power_law_1.01,10.466566467285157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,1,power_law_1.01,19.670816040039064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,1,power_law_1.2,0.12567039728164672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,1,power_law_1.2,0.2008768081665039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,1,power_law_1.2,0.23838720321655274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,1,power_law_1.2,0.39212799072265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,1,power_law_1.2,0.6468095779418945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,1,power_law_1.2,0.9614080429077149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,1,power_law_1.2,1.1368191719055176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,1,power_law_1.2,1.5849408149719237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,1,power_law_1.2,1.7010112762451173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,1,power_law_1.2,1.7345216751098633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,1,power_law_1.2,1.832102394104004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,1,power_law_1.2,1.95994873046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,1,power_law_1.2,2.003001594543457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,1,power_law_1.2,2.0943103790283204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,1,power_law_1.2,2.1334016799926756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,1,power_law_1.2,2.2039552688598634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,1,power_law_1.2,2.3060735702514648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,1,power_law_1.2,2.440108871459961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,1,power_law_1.2,2.5786943435668945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,1,power_law_1.2,0.09241600036621093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,1,power_law_1.2,2.8484224319458007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,1,power_law_1.2,0.12903679609298707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,1,power_law_1.2,3.0565248489379884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,1,power_law_1.2,0.16484479904174804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,1,power_law_1.2,3.545542526245117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,1,power_law_1.2,0.24679040908813477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,1,power_law_1.2,3.9380161285400392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,1,power_law_1.2,0.3667392015457153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,1,power_law_1.2,0.5266623973846436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,1,power_law_1.2,4.792012786865234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,1,power_law_1.2,0.6327231884002685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,1,power_law_1.2,5.577376174926758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,1,power_law_1.2,0.8960063934326172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,1,power_law_1.2,7.2997184753417965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,1,power_law_1.2,0.9196928024291993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,1,power_law_1.2,0.9607744216918945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,1,power_law_1.2,8.872652435302735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,1,power_law_1.2,1.0013567924499511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,1,power_law_1.2,10.504953765869141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,1,power_law_1.2,1.0356800079345703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,1,power_law_1.2,1.0932415962219237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,1,power_law_1.2,15.430873107910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,1,power_law_1.2,1.1478464126586914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,1,power_law_1.2,1.180339241027832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,1,power_law_1.2,28.442776489257813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,1,power_law_1.2,1.206777572631836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,1,power_law_1.2,1.2756799697875976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,1,power_law_1.2,1.369593620300293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,1,power_law_1.2,1.4668352127075195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,1,power_law_1.2,1.6168703079223632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,1,power_law_1.2,1.7884735107421874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,1,power_law_1.2,2.1247167587280273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,1,power_law_1.2,2.418284797668457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,1,power_law_1.2,2.997100830078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,1,power_law_1.2,0.13393280506134034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,1,power_law_1.2,0.1682752013206482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,1,power_law_1.2,3.5625919342041015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,4,balanced,0.16243199507395426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,1,power_law_1.2,0.1991744041442871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,4,balanced,0.16698666413625082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,1,power_law_1.2,4.720428848266602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,4,balanced,0.2315573294957479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,1,power_law_1.2,0.28421759605407715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,4,balanced,0.23248000939687094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,1,power_law_1.2,5.906399917602539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,4,balanced,0.22111467520395914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,1,power_law_1.2,0.43378558158874514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,4,balanced,0.22780267397562662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,1,power_law_1.2,7.0329345703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,1,power_law_1.2,0.5714879989624023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,4,balanced,0.24410667022069296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,4,balanced,0.2214240034421285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,1,power_law_1.2,0.6492800235748291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,1,power_law_1.2,10.437939453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,4,balanced,0.2087093393007914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,4,balanced,0.22312533855438232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,1,power_law_1.2,0.9107583999633789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,4,balanced,0.20703999201456705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,1,power_law_1.2,0.9335871696472168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,4,balanced,0.19088000059127808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,1,power_law_1.2,19.624205017089842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,4,balanced,0.2121493419011434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,1,power_law_1.2,0.9609087944030762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,4,balanced,0.19673067331314087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,1,power_law_1.2,1.0218303680419922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,4,balanced,0.18152000506718954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,4,balanced,0.1766080061594645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,1,power_law_1.2,1.0653887748718263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,4,balanced,0.17670400937398276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,1,power_law_1.2,1.0831744194030761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,4,balanced,0.28563199440638226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,1,power_law_1.2,1.1395520210266112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,4,balanced,0.25360000133514404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,1,power_law_1.2,1.1715264320373535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,4,balanced,0.3804853359858195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,1,power_law_1.2,1.2062399864196778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,4,balanced,0.4572426478068034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,1,power_law_1.2,1.2617216110229492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,4,balanced,0.6582719882329305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,1,power_law_1.2,1.3404224395751954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,4,balanced,0.857642650604248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.2,1.439743995666504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.2,1.5981247901916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,4,balanced,1.2354880174001057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.2,1.7243263244628906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,4,balanced,1.5910773277282715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.2,2.0647872924804687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.2,2.371609687805176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.2,2.9773632049560548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,4,balanced,2.254570643107096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,1,power_law_1.2,3.561676788330078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,1,power_law_1.2,4.713100814819336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,4,balanced,2.923647880554199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,1,power_law_1.2,5.824095916748047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,1,power_law_1.2,7.057849884033203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,1,power_law_1.2,10.398802947998046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,4,balanced,3.4851840337117515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,1,power_law_1.2,19.560345458984376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,4,balanced,5.216357231140137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,4,balanced,0.0990826686223348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,4,balanced,0.10982400178909302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,4,balanced,0.1353386640548706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,4,balanced,0.13914666573206583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,4,balanced,0.13352533181508383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,4,balanced,0.13613866766293845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,4,balanced,0.14808533589045206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,4,balanced,0.12665067116419473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,4,balanced,0.1340053379535675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,4,balanced,0.14046399792035422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,4,balanced,9.687119801839193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,4,balanced,0.13640000422795615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,4,balanced,0.13238400220870972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,4,balanced,0.1279146671295166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,4,balanced,0.11983999609947205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,4,balanced,0.12219732999801636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,4,balanced,0.12013866504033406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,4,balanced,0.12185066938400269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,4,balanced,0.18163732687632242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,4,balanced,0.16320000092188516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,4,balanced,0.2260800004005432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,4,balanced,0.26410667101542157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,4,balanced,0.08920533458391826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,4,balanced,0.08944533268610637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,4,balanced,0.36690131823221844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,4,balanced,0.13568533460299173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,4,balanced,0.1222879985968272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,4,balanced,0.12147733569145203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,4,balanced,0.46557335058848065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,4,balanced,0.12506666779518127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,4,balanced,0.12641599774360657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,4,balanced,0.1290079951286316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,4,balanced,0.6552693446477255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,4,balanced,0.12944533427556357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,4,balanced,0.12533866365750632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,4,balanced,0.11622933546702068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,4,balanced,0.8383626937866211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,4,balanced,0.12124799688657124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,4,balanced,0.11643733580907185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,4,balanced,0.10493333141009013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,4,balanced,0.10859200358390808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,4,balanced,0.10152000188827515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,4,balanced,1.2042720317840576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,4,balanced,0.09698667128880818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,4,balanced,0.1581546664237976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,4,balanced,0.13251200318336487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,4,balanced,0.17118932803471884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,4,balanced,1.5692960421244304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,4,balanced,0.20775467157363892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,4,balanced,0.2826559940973918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,4,balanced,0.3529386520385742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,4,balanced,1.9264267285664876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,4,balanced,0.5089919964472452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,4,balanced,0.6499946514765421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,4,balanced,0.9530346393585205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,4,balanced,3.000613212585449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,4,balanced,1.2558986345926921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,4,balanced,0.03426666557788849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,4,balanced,0.03368533402681351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,4,balanced,0.033258666594823204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,4,balanced,1.5394293467203777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,4,balanced,0.03909866760174433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,4,balanced,0.03827200084924698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,4,balanced,0.04045333216587702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,4,balanced,0.06302399933338165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,4,balanced,0.046495998899141945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,4,balanced,0.08591999610265096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,4,balanced,0.0863200028737386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,4,balanced,0.08545066912968953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,4,balanced,2.446986675262451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,4,balanced,5.728895823160808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,4,balanced,0.10342933734258015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,4,balanced,0.09452799956003825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,4,balanced,0.0703413337469101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,4,balanced,0.10894933342933655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,4,power_law_1.01,0.1962623953819275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,4,balanced,0.08413333694140117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,4,power_law_1.01,0.2204927921295166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,4,balanced,0.09456533193588257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,4,balanced,0.13363200426101685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,4,balanced,0.158053328593572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,4,balanced,0.22103999058405557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,4,balanced,4.816287994384766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,4,balanced,0.2789546648661296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,4,balanced,0.39715198675791424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,4,balanced,0.5017653306325277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,4,balanced,0.740938663482666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,4,balanced,0.9657386938730875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,4,balanced,1.4183519681294758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,4,balanced,1.8629813194274902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,4,balanced,2.3270719846089682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,4,balanced,3.6618614196777344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,4,balanced,7.2844797770182295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,4,power_law_1.01,0.11994240283966065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,4,power_law_1.01,0.11559679508209228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,4,power_law_1.01,0.11548800468444824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,4,power_law_1.01,0.12942080497741698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,4,power_law_1.01,0.13262720108032228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,4,power_law_1.01,0.13646080493927001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,4,power_law_1.01,0.13255679607391357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,4,power_law_1.01,0.13218560218811035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,4,power_law_1.01,0.13351680040359498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,4,power_law_1.01,0.1298815965652466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,4,power_law_1.01,0.12218879461288452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,4,power_law_1.01,0.13712639808654786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,4,power_law_1.01,0.15002880096435547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,4,power_law_1.01,0.1415552020072937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,4,power_law_1.01,0.16316159963607788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,4,power_law_1.01,0.16407040357589722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,4,power_law_1.01,0.21368958950042724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,4,power_law_1.01,0.2677504062652588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,4,power_law_1.01,0.10687359571456909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,4,power_law_1.01,0.3337280035018921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,4,power_law_1.01,0.11973760128021241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,4,power_law_1.01,0.4063231945037842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,4,power_law_1.01,0.11937919855117798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,4,power_law_1.01,0.4953279972076416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,4,power_law_1.01,0.12559360265731812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,4,power_law_1.01,0.7604032039642334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,4,power_law_1.01,0.12405760288238525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,4,power_law_1.01,1.0454848289489747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,4,power_law_1.01,0.1270848035812378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,4,power_law_1.01,1.5346176147460937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,4,power_law_1.01,0.12117120027542114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,4,power_law_1.01,2.1964672088623045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,4,power_law_1.01,0.12293119430541992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,4,power_law_1.01,2.670803260803223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,4,power_law_1.01,0.11992319822311401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,4,power_law_1.01,0.1198464035987854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,4,power_law_1.01,3.5306880950927733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,4,power_law_1.01,0.11251200437545776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,4,power_law_1.01,4.499692916870117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,4,power_law_1.01,0.12114559412002564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,4,power_law_1.01,6.471199798583984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,4,power_law_1.01,0.127455997467041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,4,power_law_1.01,0.13292800188064574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,4,power_law_1.01,13.361708068847657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,4,power_law_1.01,0.15022720098495485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,4,power_law_1.01,0.13886719942092896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,4,power_law_1.01,0.1804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,4,power_law_1.01,0.22035200595855714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.01,0.24131200313568116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.01,0.3725951910018921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.01,0.43157758712768557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.01,0.6770112037658691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.01,0.7172095775604248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.01,1.1715392112731933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.01,1.5209856033325195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.01,2.031711959838867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.01,3.05281925201416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,4,power_law_1.01,4.336595153808593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,4,power_law_1.01,6.7794044494628904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,4,power_law_1.01,10.548818969726563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,4,power_law_1.01,0.03163520097732544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,4,power_law_1.01,0.20702080726623534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,4,power_law_1.01,0.03227519989013672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,4,power_law_1.01,0.2357759952545166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,4,power_law_1.01,0.030623999238014222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,4,power_law_1.01,0.22708480358123778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,4,power_law_1.01,0.036556801199913024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,4,power_law_1.01,0.21018240451812745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,4,power_law_1.01,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,4,power_law_1.01,0.21362559795379638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,4,power_law_1.01,0.2093183994293213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,4,power_law_1.01,0.0381056010723114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,4,power_law_1.01,0.20351359844207764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,4,power_law_1.01,0.05867519974708557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,4,power_law_1.01,0.20753281116485595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,4,power_law_1.01,0.04334079921245575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,4,power_law_1.01,0.1954624056816101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,4,power_law_1.01,0.08431360125541687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,4,power_law_1.01,0.0832319974899292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,4,power_law_1.01,0.22400639057159424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,4,power_law_1.01,0.24109439849853515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,4,power_law_1.01,0.0844223976135254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,4,power_law_1.01,0.23301761150360106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,4,power_law_1.01,0.09903360009193421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,4,power_law_1.01,0.3188352108001709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,4,power_law_1.01,0.09287040233612061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,4,power_law_1.01,0.06750720143318176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,4,power_law_1.01,0.29404799938201903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,4,power_law_1.01,0.10875519514083862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,4,power_law_1.01,0.39107840061187743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,4,power_law_1.01,0.08252800107002259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,4,power_law_1.01,0.5002751827239991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,4,power_law_1.01,0.09174399971961975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,4,power_law_1.01,0.5279295921325684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,4,power_law_1.01,0.13144960403442382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,4,power_law_1.01,0.743123197555542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.01,0.15489280223846436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,4,power_law_1.01,0.9212479591369629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.01,0.21889278888702393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,4,power_law_1.01,1.2998847961425781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.01,0.28061439990997317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,4,power_law_1.01,2.1033920288085937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.01,0.3970047950744629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,4,power_law_1.01,2.7263168334960937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.01,0.5034687995910645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,4,power_law_1.01,3.484832000732422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.01,0.7371520042419434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,4,power_law_1.01,4.240998458862305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.01,0.9643008232116699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,4,power_law_1.01,7.991577911376953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.01,1.4104512214660645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.01,1.8588735580444335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,4,power_law_1.01,6.753017425537109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,4,power_law_1.01,2.3076927185058596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,4,power_law_1.01,13.665242004394532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,4,power_law_1.01,3.6678398132324217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,4,power_law_1.01,7.281388854980468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,4,power_law_1.01,24.869894409179686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,4,power_law_1.2,0.19685120582580568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,4,power_law_1.2,0.20185599327087403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,4,power_law_1.2,0.2277440071105957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,4,power_law_1.2,0.22536320686340333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,4,power_law_1.2,0.215283203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,4,power_law_1.2,0.22707839012145997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,4,power_law_1.2,0.22155520915985108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,4,power_law_1.2,0.21461760997772217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,4,power_law_1.2,0.19731839895248413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,4,power_law_1.2,0.2034048080444336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,4,power_law_1.2,0.1943935990333557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,4,power_law_1.2,0.21841919422149658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,4,power_law_1.2,0.21646718978881835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,4,power_law_1.2,0.2388223886489868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,4,power_law_1.2,0.24518399238586425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,4,power_law_1.2,0.1078976035118103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,4,power_law_1.2,0.3147007942199707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,4,power_law_1.2,0.13586560487747193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,4,power_law_1.2,0.41166081428527834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,4,power_law_1.2,0.13057279586791992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,4,power_law_1.2,0.46645121574401854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,4,power_law_1.2,0.1296447992324829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,4,power_law_1.2,0.5880767822265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,4,power_law_1.2,0.14149760007858275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,4,power_law_1.2,0.8191679954528809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,4,power_law_1.2,0.1362048029899597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,4,power_law_1.2,1.0397055625915528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,4,power_law_1.2,0.1326464056968689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,4,power_law_1.2,1.435206413269043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,4,power_law_1.2,0.13527040481567382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,4,power_law_1.2,2.1075456619262694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,4,power_law_1.2,0.13338240385055541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,4,power_law_1.2,3.098150444030762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,4,power_law_1.2,0.1285696029663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,4,power_law_1.2,4.371615982055664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,4,power_law_1.2,0.12811520099639892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,4,power_law_1.2,0.13780479431152343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,4,power_law_1.2,4.708729553222656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,4,power_law_1.2,0.14380799531936644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,4,power_law_1.2,6.6827842712402346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,4,power_law_1.2,0.14270080327987672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,4,power_law_1.2,0.18504960536956788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,4,power_law_1.2,8.745120239257812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,4,power_law_1.2,0.19179519414901733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,4,power_law_1.2,11.773868560791016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,4,power_law_1.2,0.21948161125183105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,4,power_law_1.2,0.30796160697937014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,4,power_law_1.2,0.3833343982696533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,4,power_law_1.2,27.271096801757814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,4,power_law_1.2,0.48857598304748534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,4,power_law_1.2,0.5525119781494141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,4,power_law_1.2,0.7807231903076172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,4,power_law_1.2,0.9195455551147461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,4,power_law_1.2,1.5858431816101075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,4,power_law_1.2,1.9799936294555665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,4,power_law_1.2,3.0279872894287108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,4,power_law_1.2,4.089606475830078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,4,power_law_1.2,5.25665283203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,4,power_law_1.2,8.012505340576173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,4,power_law_1.2,14.906643676757813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,4,power_law_1.2,0.09679999947547913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,4,power_law_1.2,0.03314560055732727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,4,power_law_1.2,0.11573120355606079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,4,power_law_1.2,0.03265919983386993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,4,power_law_1.2,0.11667200326919555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,4,power_law_1.2,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,4,power_law_1.2,0.12891520261764527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,4,power_law_1.2,0.037190398573875426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,4,power_law_1.2,0.125273597240448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,4,power_law_1.2,0.037990400195121767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,4,power_law_1.2,0.12396800518035889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,4,power_law_1.2,0.03808000087738037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,4,power_law_1.2,0.12028160095214843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,4,power_law_1.2,0.05861120223999024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,4,power_law_1.2,0.1140544056892395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,4,power_law_1.2,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,4,power_law_1.2,0.11959680318832397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,4,power_law_1.2,0.08636800050735474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,4,power_law_1.2,0.11648000478744507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,4,power_law_1.2,0.08647680282592773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,4,power_law_1.2,0.10374399423599243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,4,power_law_1.2,0.08459519743919372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,4,power_law_1.2,0.13749120235443116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,4,power_law_1.2,0.09420160055160523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,4,power_law_1.2,0.14722559452056885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,4,power_law_1.2,0.08807039856910706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,4,power_law_1.2,0.11347839832305909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,4,power_law_1.2,0.06840959787368775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,4,power_law_1.2,0.1665536046028137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,4,power_law_1.2,0.10535680055618286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,4,power_law_1.2,0.15445760488510132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,4,power_law_1.2,0.08224639892578126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,4,power_law_1.2,0.1916864037513733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,4,power_law_1.2,0.09192960262298584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,4,power_law_1.2,0.22176001071929932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,4,power_law_1.2,0.1299456000328064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.2,0.2873408079147339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.2,0.15432319641113282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.2,0.3535167932510376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.2,0.22030720710754395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.2,0.4240896224975586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.2,0.27936639785766604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.2,0.39185919761657717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.2,0.6813055992126464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.2,0.5022208213806152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.2,0.9033791542053222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.2,0.7376768112182617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.2,1.3406911849975587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.2,0.9627776145935059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.2,1.5299327850341797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.2,1.4185791969299317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.2,2.444460868835449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.2,1.8572416305541992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.2,3.6237953186035154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,4,power_law_1.2,2.327257537841797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,4,power_law_1.2,3.9176704406738283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,4,power_law_1.2,3.6649921417236326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,4,power_law_1.2,7.486796569824219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,4,power_law_1.2,7.2829948425292965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,4,power_law_1.2,14.595840454101562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,2,balanced,0.10078932841618855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,2,balanced,0.141157329082489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,2,balanced,0.1879840095837911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,2,balanced,0.17882666985193887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,2,balanced,0.19100266695022583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,2,balanced,0.19355199734369913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,2,balanced,0.19987734158833823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,2,balanced,0.19244267543156943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,2,balanced,0.18880534172058105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,2,balanced,0.18870399395624796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,2,balanced,0.18039466937383017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,2,balanced,0.06678399940331776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,2,balanced,0.185754656791687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,2,balanced,0.08986666798591614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,2,balanced,0.17294400930404663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,2,balanced,0.12730666995048523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,2,balanced,0.1639199952284495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,2,balanced,0.11755733688672383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,2,balanced,0.16403200229008993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,2,balanced,0.11475200454394023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,2,balanced,0.15822399655977884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,2,balanced,0.11101866761843364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,2,balanced,0.13064000010490417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,2,balanced,0.16259732842445374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,2,balanced,0.11505599816640218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,2,balanced,0.2560746669769287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,2,balanced,0.13212800025939941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,2,balanced,0.24826133251190186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,2,balanced,0.1239306628704071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,2,balanced,0.11715199549992879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,2,balanced,0.33779199918111164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,2,balanced,0.10729599992434184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,2,balanced,0.11245333154996236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,2,balanced,0.41229867935180664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,2,balanced,0.1086133321126302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,2,balanced,0.5683146715164185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,2,balanced,0.11327999830245972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,2,balanced,0.1127306620279948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,2,balanced,0.7048746744791666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,2,balanced,0.11581866939862569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,2,balanced,0.1685333251953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,2,balanced,0.15570132931073508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,2,balanced,0.9927840232849121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,2,balanced,0.20611733198165894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,2,balanced,0.2468000054359436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,2,balanced,1.2665546735127766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,2,balanced,0.3402239878972371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,2,balanced,0.42215466499328613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,2,balanced,1.8255680402119954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,2,balanced,0.6077280044555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,2,balanced,0.7886559963226318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,2,balanced,0.06480533381303151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,2,balanced,2.383061408996582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,2,balanced,0.023904000719388325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,2,balanced,0.08010666569073994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,2,balanced,0.02456533412138621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,2,balanced,0.13036800424257913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,2,balanced,1.153274695078532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,2,balanced,0.024559999505678814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,2,balanced,0.11807466546694438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,2,balanced,0.028783999383449554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,2,balanced,0.11675199866294861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,2,balanced,0.030784000953038532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,2,balanced,0.11639466881752014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,2,balanced,0.032560000816980995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,2,balanced,0.1237546702226003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,2,balanced,2.944016138712565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,2,balanced,1.5218186378479004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,2,balanced,0.03180266668399175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,2,balanced,0.12229866782824199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,2,balanced,0.12009599804878235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,2,balanced,0.0323786661028862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,2,balanced,0.12943466504414877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,2,balanced,0.04064533362785975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,2,balanced,0.11132799585660298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,2,balanced,0.03544000039498011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,2,balanced,0.11916800340016682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,2,balanced,0.036576000352700554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,2,balanced,1.8801973660786946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,2,balanced,0.11212799946467082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,2,balanced,0.07566933333873749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,2,balanced,0.10421866178512573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,2,balanced,0.07605866591135661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,2,balanced,0.10046399633089702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,2,balanced,0.0776800016562144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,2,balanced,0.10345600048700969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,2,balanced,0.0663679987192154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,2,balanced,4.574901262919108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,2,balanced,0.09822400410970052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,2,balanced,0.057802667220433555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,2,balanced,0.16517866651217142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,2,balanced,0.05919999877611796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,2,balanced,2.945215861002604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,2,balanced,0.1360426644484202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,2,balanced,0.08245866497357686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,2,balanced,0.0937653382619222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,2,balanced,0.17634665966033936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,2,balanced,0.13077333569526672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,2,balanced,0.2108853260676066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,2,balanced,0.1618133286635081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,2,balanced,0.2898240089416504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,2,balanced,0.2276853322982788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,2,balanced,0.3685386578241984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,2,balanced,0.28589866558710736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,2,balanced,0.5181653499603271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,2,balanced,0.4151093165079753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,2,balanced,0.6695466836293539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,2,balanced,5.777930577596028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,2,balanced,9.130874633789062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,2,balanced,0.5357600053151449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,2,balanced,0.9802666505177816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,2,balanced,0.7805333137512207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,2,balanced,1.2872107028961182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,2,balanced,1.0283253192901611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,2,balanced,1.5994826952616374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,2,balanced,1.275264024734497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,2,balanced,2.534272034962972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,2,balanced,2.012394587198893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,2,balanced,5.0992692311604815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,2,balanced,3.9882987340291343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,2,power_law_1.01,0.12925440073013306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,2,power_law_1.01,0.08318719863891602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,2,power_law_1.01,0.1385151982307434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,2,power_law_1.01,0.08492799997329711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,2,power_law_1.01,0.141593599319458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,2,power_law_1.01,0.09699199795722961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,2,power_law_1.01,0.1749567985534668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,2,power_law_1.01,0.1224511981010437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,2,power_law_1.01,0.19530240297317505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,2,power_law_1.01,0.11592960357666016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,2,power_law_1.01,0.19349759817123413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,2,power_law_1.01,0.11931519508361817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,2,power_law_1.01,0.19352960586547852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,2,power_law_1.01,0.11907839775085449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,2,power_law_1.01,0.19086079597473143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,2,power_law_1.01,0.11536639928817749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,2,power_law_1.01,0.19173120260238646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,2,power_law_1.01,0.12482559680938721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,2,power_law_1.01,0.18888959884643555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,2,power_law_1.01,0.12001919746398926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,2,power_law_1.01,0.17879040241241456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,2,power_law_1.01,0.11326719522476196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,2,power_law_1.01,0.21000320911407472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,2,power_law_1.01,0.11829760074615478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,2,power_law_1.01,0.22739200592041015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,2,power_law_1.01,0.13712639808654786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,2,power_law_1.01,0.2261888027191162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,2,power_law_1.01,0.13008639812469483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,2,power_law_1.01,0.2479935884475708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,2,power_law_1.01,0.15648000240325927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,2,power_law_1.01,0.2483520030975342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,2,power_law_1.01,0.157478404045105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,2,power_law_1.01,0.26267518997192385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,2,power_law_1.01,0.16836479902267457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,2,power_law_1.01,0.3564543962478638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,2,power_law_1.01,0.19560960531234742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,2,power_law_1.01,0.4029183864593506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,2,power_law_1.01,0.23736960887908937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,2,power_law_1.01,0.510975980758667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,2,power_law_1.01,0.30483839511871336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,2,power_law_1.01,0.3630527973175049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,2,power_law_1.01,0.5715199947357178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,2,power_law_1.01,0.5121984004974365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,2,power_law_1.01,0.7753087997436523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,2,power_law_1.01,0.547814416885376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,2,power_law_1.01,1.0878144264221192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,2,power_law_1.01,0.9747776031494141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,2,power_law_1.01,1.6088640213012695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,2,power_law_1.01,0.07365760207176208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,2,power_law_1.01,1.307040023803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,2,power_law_1.01,1.796748733520508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,2,power_law_1.01,0.08439040184020996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,2,power_law_1.01,1.725209617614746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,2,power_law_1.01,3.000160026550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,2,power_law_1.01,0.09182080030441284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,2,power_law_1.01,2.267411231994629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,2,power_law_1.01,3.442483139038086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,2,power_law_1.01,0.12685439586639405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,2,power_law_1.01,3.1229375839233398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,2,power_law_1.01,0.11934720277786255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,2,power_law_1.01,4.232595062255859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,2,power_law_1.01,0.12289919853210449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,2,power_law_1.01,4.26126708984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,2,power_law_1.01,7.5259651184082035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,2,power_law_1.01,0.1232640027999878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,2,power_law_1.01,9.319750213623047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,2,power_law_1.01,0.11951999664306641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,2,power_law_1.01,0.023244799673557283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,2,power_law_1.01,14.845120239257813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,2,power_law_1.01,0.11864320039749146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,2,power_law_1.01,0.023654399812221526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,2,power_law_1.01,0.11991039514541627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,2,power_law_1.01,0.022963200509548188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,2,power_law_1.01,0.11342079639434814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,2,power_law_1.01,0.027251198887825012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,2,power_law_1.01,0.12501120567321777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,2,power_law_1.01,0.030323201417922975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,2,power_law_1.01,0.1245695948600769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,2,power_law_1.01,0.03121280074119568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,2,power_law_1.01,0.12948479652404785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,2,power_law_1.01,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,2,power_law_1.01,0.147705602645874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,2,power_law_1.01,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,2,power_law_1.01,0.13628159761428832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,2,power_law_1.01,0.03797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,2,power_law_1.01,0.16218240261077882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,2,power_law_1.01,0.03542400002479553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,2,power_law_1.01,0.1832703948020935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,2,power_law_1.01,0.03555200099945068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.01,0.20532479286193847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,2,power_law_1.01,0.07414399981498718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,2,power_law_1.01,0.07409279942512512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.01,0.25333120822906496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,2,power_law_1.01,0.07631999850273133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.01,0.33517439365386964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,2,power_law_1.01,0.06355839967727661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.01,0.4507391929626465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,2,power_law_1.01,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.01,0.5386112213134766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,2,power_law_1.01,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.01,0.7897088050842285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,2,power_law_1.01,0.08201599717140198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.01,1.1291199684143067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.01,0.09257599711418152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.01,1.3347583770751954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.01,0.13064320087432862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.01,1.8395135879516602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.01,0.16307200193405152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,2,power_law_1.01,2.841721534729004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.01,0.22645120620727538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,2,power_law_1.01,4.377791976928711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.01,0.2869375944137573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.01,0.4146240234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,2,power_law_1.01,8.326681518554688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.01,0.5391104221343994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.01,0.7796864032745361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.01,1.0294079780578613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,2,power_law_1.01,1.2748543739318847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,2,power_law_1.01,2.0124671936035154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,2,power_law_1.01,3.9885185241699217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,2,power_law_1.2,0.12017279863357544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,2,power_law_1.2,0.07764480113983155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,2,power_law_1.2,0.12830719947814942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,2,power_law_1.2,0.08300799727439881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,2,power_law_1.2,0.13687679767608643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,2,power_law_1.2,0.09631999731063842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,2,power_law_1.2,0.17788159847259521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,2,power_law_1.2,0.11938560009002686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,2,power_law_1.2,0.1958143949508667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,2,power_law_1.2,0.11233279705047608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,2,power_law_1.2,0.1936959981918335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,2,power_law_1.2,0.12495360374450684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,2,power_law_1.2,0.19027199745178222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,2,power_law_1.2,0.11745280027389526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,2,power_law_1.2,0.19158400297164918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,2,power_law_1.2,0.12061439752578736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,2,power_law_1.2,0.19059200286865235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,2,power_law_1.2,0.122489595413208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,2,power_law_1.2,0.1928063988685608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,2,power_law_1.2,0.11908479928970336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,2,power_law_1.2,0.18160640001296996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,2,power_law_1.2,0.11466879844665527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,2,power_law_1.2,0.18996479511260986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,2,power_law_1.2,0.13494399785995484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,2,power_law_1.2,0.23482239246368408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,2,power_law_1.2,0.14497920274734497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,2,power_law_1.2,0.22142720222473145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,2,power_law_1.2,0.14684799909591675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,2,power_law_1.2,0.24932479858398438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,2,power_law_1.2,0.1463487982749939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,2,power_law_1.2,0.2621504068374634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,2,power_law_1.2,0.15491199493408203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,2,power_law_1.2,0.28548479080200195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,2,power_law_1.2,0.1745344042778015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,2,power_law_1.2,0.348639988899231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,2,power_law_1.2,0.19738880395889283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,2,power_law_1.2,0.23400321006774902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,2,power_law_1.2,0.4417280197143555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,2,power_law_1.2,0.5180287837982178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,2,power_law_1.2,0.30064640045166013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,2,power_law_1.2,0.6733248233795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,2,power_law_1.2,0.38172159194946287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,2,power_law_1.2,0.07870720028877258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,2,power_law_1.2,0.847878360748291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,2,power_law_1.2,0.5507264137268066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,2,power_law_1.2,0.07852799892425537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,2,power_law_1.2,1.0285311698913575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,2,power_law_1.2,0.740499210357666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,2,power_law_1.2,0.09222400188446045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,2,power_law_1.2,1.5819775581359863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,2,power_law_1.2,1.0324224472045898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,2,power_law_1.2,0.12439039945602418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,2,power_law_1.2,1.1877375602722169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,2,power_law_1.2,2.0534400939941406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,2,power_law_1.2,0.11761280298233032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,2,power_law_1.2,1.6494976043701173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,2,power_law_1.2,2.687167930603027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,2,power_law_1.2,0.12345600128173828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,2,power_law_1.2,2.168550491333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,2,power_law_1.2,0.11715199947357177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,2,power_law_1.2,3.5455936431884765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,2,power_law_1.2,0.12168320417404174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,2,power_law_1.2,3.0243839263916015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,2,power_law_1.2,4.309766387939453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,2,power_law_1.2,0.11802879571914673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,2,power_law_1.2,5.371622467041016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,2,power_law_1.2,7.163353729248047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,2,power_law_1.2,0.11693439483642579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,2,power_law_1.2,0.022988800704479218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,2,power_law_1.2,0.11554559469223022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,2,power_law_1.2,8.993702697753907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,2,power_law_1.2,0.024255999922752382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,2,power_law_1.2,17.288114929199217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,2,power_law_1.2,0.11473920345306396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,2,power_law_1.2,0.02369280010461807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,2,power_law_1.2,0.12277760505676269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,2,power_law_1.2,0.027667200565338133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,2,power_law_1.2,0.12743680477142333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,2,power_law_1.2,0.03177599906921387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,2,power_law_1.2,0.14657280445098878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,2,power_law_1.2,0.031455999612808226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,2,power_law_1.2,0.14394240379333495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,2,power_law_1.2,0.030899199843406677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,2,power_law_1.2,0.15700479745864868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,2,power_law_1.2,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,2,power_law_1.2,0.1942720055580139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,2,power_law_1.2,0.04085760116577149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.2,0.2113408088684082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,2,power_law_1.2,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.2,0.2664448022842407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,2,power_law_1.2,0.03617280125617981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.2,0.3052031993865967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,2,power_law_1.2,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.2,0.47868800163269043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,2,power_law_1.2,0.07490559816360473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.2,0.5268735885620117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,2,power_law_1.2,0.07774080038070678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.2,0.7615744113922119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,2,power_law_1.2,0.0648576021194458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.2,1.0816384315490724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,2,power_law_1.2,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.2,1.516703987121582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,2,power_law_1.2,0.05864959955215454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.2,1.9521919250488282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,2,power_law_1.2,0.08138880133628845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.2,0.09221119880676269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,2,power_law_1.2,2.5000768661499024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.2,0.12987519502639772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,2,power_law_1.2,4.260684967041016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.2,0.16341760158538818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.2,0.22768640518188477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,2,power_law_1.2,7.846931457519531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.2,0.28672640323638915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.2,0.4139840126037598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.2,0.536902379989624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.2,0.7836351871490479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.2,1.0297856330871582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,2,power_law_1.2,1.2798912048339843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,2,power_law_1.2,2.106912040710449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,2,power_law_1.2,3.990438461303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,1,balanced,0.08467732866605122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,1,balanced,0.05515733361244202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,1,balanced,0.1039626697699229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,1,balanced,0.07322666545708974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,1,balanced,0.1719520092010498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,1,balanced,0.12680000066757202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,1,balanced,0.17392534017562866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,1,balanced,0.10752000411351521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,1,balanced,0.1746293306350708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,1,balanced,0.17390932639439902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,1,balanced,0.10224533081054688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,1,balanced,0.19333332777023315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,1,balanced,0.11311999956766765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,1,balanced,0.18127467234929404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,1,balanced,0.1253493328889211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,1,balanced,0.1853920022646586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,1,balanced,0.11769066254297893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,1,balanced,0.17691733439763388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,1,balanced,0.11452800035476685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,1,balanced,0.1755946675936381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,1,balanced,0.1095199982325236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,1,balanced,0.1841920018196106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,1,balanced,0.11775466799736023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,1,balanced,0.16849066813786825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,1,balanced,0.1112000048160553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,1,balanced,0.1606933375199636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,1,balanced,0.1095306674639384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,1,balanced,0.16217066844304404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,1,balanced,0.10485333204269409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,1,balanced,0.16197333733240762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,1,balanced,0.11269866426785786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,1,balanced,0.16716800133387247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,1,balanced,0.11351466178894043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,1,balanced,0.2560746669769287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,1,balanced,0.11760000387827556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,1,balanced,0.23710399866104126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,1,balanced,0.16245333353678384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,1,balanced,0.30929599205652875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,1,balanced,0.16153066356976828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,1,balanced,0.2042400042215983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,1,balanced,0.3712213436762492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,1,balanced,0.0629066675901413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,1,balanced,0.07793599863847096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,1,balanced,0.2469386657079061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,1,balanced,0.5163466533025106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,1,balanced,0.12735999623934427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,1,balanced,0.12528533736864725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,1,balanced,0.342410683631897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,1,balanced,0.6448053518931071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,1,balanced,0.11756267150243123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,1,balanced,0.11015466849009196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,1,balanced,0.43086934089660645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,1,balanced,0.9458719889322916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,1,balanced,0.12488533059755962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,1,balanced,0.10938133796056111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,1,balanced,0.021589333812395733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,1,balanced,0.624837319056193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,1,balanced,0.11780800422032674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,1,balanced,0.02316266546646754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,1,balanced,1.229967991511027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,1,balanced,0.12065600355466206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,1,balanced,0.022416000564893086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,1,balanced,0.024080000817775726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,1,balanced,0.10776533683141072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,1,balanced,0.026528000831604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,1,balanced,0.10848533113797505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,1,balanced,0.8090666929880778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,1,balanced,0.03081600119670232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,1,balanced,0.11711999773979187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,1,balanced,0.030373332401116688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,1,balanced,1.8080800374348958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,1,balanced,0.10709333419799805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,1,balanced,0.03198933353026708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,1,balanced,0.11050666371981303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,1,balanced,0.02959999938805898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,1,balanced,1.1933866341908772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,1,balanced,0.10190932949384053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,1,balanced,0.03260799994071325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,1,balanced,0.10302933057149251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,1,balanced,0.03477866699298223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,1,balanced,2.4048479398091636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,1,balanced,0.03249066571394602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,1,balanced,0.16331199804941812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,1,balanced,0.03698666642109553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,1,balanced,0.14849600195884705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,1,balanced,0.03732266773780187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,1,balanced,1.5756905873616536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,1,balanced,0.19048533837000528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,1,balanced,0.06265600025653839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,1,balanced,0.07753066718578339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,1,balanced,0.23355199893315634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,1,balanced,3.0021438598632812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,1,balanced,0.07958399752775829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,1,balanced,0.3152479926745097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,1,balanced,0.058965335289637245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,1,balanced,1.9672959645589192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,1,balanced,0.06152533491452535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,1,balanced,0.3985813458760579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,1,balanced,0.08509332935015361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,1,balanced,0.0995199978351593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,1,balanced,0.5664000113805135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,1,balanced,4.806607882181804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,1,balanced,0.13780799508094788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,1,balanced,0.7368853092193604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,1,balanced,0.17101866006851196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,1,balanced,3.088629404703776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,1,balanced,0.24433066447575888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,1,balanced,1.0824426809946697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,1,balanced,0.3137493332227071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,1,balanced,1.4251999855041504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,1,balanced,0.4530666669209798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,1,balanced,9.618213017781576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,1,balanced,0.5941760142644247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,1,balanced,6.124810536702474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,1,balanced,1.7597707112630208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,1,balanced,0.7330239613850912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,1,balanced,2.8046773274739585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,1,balanced,1.1534667015075684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,1,balanced,2.2723520596822104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,1,balanced,5.658965428670247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,1,power_law_1.01,0.08689280152320862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,1,power_law_1.01,0.05556480288505554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,1,power_law_1.01,0.09559040069580078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,1,power_law_1.01,0.063264000415802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,1,power_law_1.01,0.11397119760513305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,1,power_law_1.01,0.07616639733314515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,1,power_law_1.01,0.10354559421539307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,1,power_law_1.01,0.17040640115737915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,1,power_law_1.01,0.11337599754333497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,1,power_law_1.01,0.1807360053062439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,1,power_law_1.01,0.11665279865264892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,1,power_law_1.01,0.17941759824752807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,1,power_law_1.01,0.1220736026763916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,1,power_law_1.01,0.1843775987625122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,1,power_law_1.01,0.11572480201721191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,1,power_law_1.01,0.18936320543289184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,1,power_law_1.01,0.1144320011138916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,1,power_law_1.01,0.1843135952949524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,1,power_law_1.01,0.1083456039428711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,1,power_law_1.01,0.18320000171661377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,1,power_law_1.01,0.11459840536117553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,1,power_law_1.01,0.18038400411605834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,1,power_law_1.01,0.11578240394592285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,1,power_law_1.01,0.18410240411758422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,1,power_law_1.01,0.12038400173187255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,1,power_law_1.01,0.19000320434570311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,1,power_law_1.01,0.12700799703598023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,1,power_law_1.01,0.20395519733428955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,1,power_law_1.01,0.13445760011672975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,1,power_law_1.01,0.21376640796661378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,1,power_law_1.01,0.14031360149383545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,1,power_law_1.01,0.2285952091217041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,1,power_law_1.01,0.14687999486923217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,1,power_law_1.01,0.2431488037109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,1,power_law_1.01,0.17614079713821412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,1,power_law_1.01,0.28481919765472413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,1,power_law_1.01,0.20921599864959717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,1,power_law_1.01,0.3220223903656006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,1,power_law_1.01,0.26119680404663087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,1,power_law_1.01,0.3960576057434082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,1,power_law_1.01,0.29363839626312255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,1,power_law_1.01,0.46881279945373533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,1,power_law_1.01,0.39959681034088135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,1,power_law_1.01,0.6101632118225098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,1,power_law_1.01,0.06365439891815186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,1,power_law_1.01,0.02059520035982132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,1,power_law_1.01,0.48349437713623045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,1,power_law_1.01,0.7639103889465332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,1,power_law_1.01,0.07710080146789551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,1,power_law_1.01,0.02247679978609085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,1,power_law_1.01,0.6739520072937012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,1,power_law_1.01,1.071558380126953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,1,power_law_1.01,0.07730559706687927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,1,power_law_1.01,0.022284799814224245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,1,power_law_1.01,0.8567935943603515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,1,power_law_1.01,1.348748779296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,1,power_law_1.01,0.1269503951072693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,1,power_law_1.01,0.022579200565814972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,1,power_law_1.01,1.2385343551635741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,1,power_law_1.01,1.9272960662841796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,1,power_law_1.01,0.12008960247039795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,1,power_law_1.01,0.02653439939022064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,1,power_law_1.01,1.626924705505371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,1,power_law_1.01,0.11175040006637574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,1,power_law_1.01,2.521683120727539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,1,power_law_1.01,0.03160319924354553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,1,power_law_1.01,1.9995712280273437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,1,power_law_1.01,0.11509120464324951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,1,power_law_1.01,3.0793024063110352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,1,power_law_1.01,0.029971200227737426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,1,power_law_1.01,0.11390719413757325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,1,power_law_1.01,3.1330432891845703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,1,power_law_1.01,0.031462401151657104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,1,power_law_1.01,4.883430480957031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,1,power_law_1.01,0.11699199676513672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,1,power_law_1.01,0.029100799560546876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,1,power_law_1.01,6.134649658203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,1,power_law_1.01,0.11527040004730224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,1,power_law_1.01,9.752275085449218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,1,power_law_1.01,0.032076799869537355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,1,power_law_1.01,0.11419520378112794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,1,power_law_1.01,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,1,power_law_1.01,0.11088639497756958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,1,power_law_1.01,0.03283199965953827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,1,power_law_1.01,0.11781120300292969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,1,power_law_1.01,0.034720000624656674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,1,power_law_1.01,0.11385600566864014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,1,power_law_1.01,0.03758719861507416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,1,power_law_1.01,0.13746559619903564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,1,power_law_1.01,0.06172159910202026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,1,power_law_1.01,0.13524479866027833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,1,power_law_1.01,0.07646080255508422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,1,power_law_1.01,0.14780160188674926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,1,power_law_1.01,0.07959679961204529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,1,power_law_1.01,0.16788480281829835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,1,power_law_1.01,0.05815039873123169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.01,0.18893439769744874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.01,0.05975040197372437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.01,0.2201024055480957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.01,0.08517119884490967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.01,0.2752255916595459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.01,0.10126719474792481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.01,0.13671679496765138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.01,0.35694079399108886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.01,0.1703552007675171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.01,0.4401408195495605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.01,0.2423680067062378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.01,0.6101439952850342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.01,0.3118272066116333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.01,0.7631360054016113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.01,0.4519680023193359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.01,1.1054656028747558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.01,0.5934912204742432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.01,1.4441023826599122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.01,0.732147216796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.01,1.787379264831543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,1,power_law_1.01,1.1536640167236327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,1,power_law_1.01,2.8156160354614257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,1,power_law_1.01,2.2739519119262694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,1,power_law_1.01,5.5772544860839846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,1,power_law_1.2,0.054016000032424925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,1,power_law_1.2,0.07381759881973267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,1,power_law_1.2,0.061504000425338747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,1,power_law_1.2,0.09411200284957885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,1,power_law_1.2,0.06876800060272217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,1,power_law_1.2,0.11172480583190918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,1,power_law_1.2,0.10066560506820679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,1,power_law_1.2,0.17556480169296265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,1,power_law_1.2,0.11046400070190429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,1,power_law_1.2,0.1799936056137085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,1,power_law_1.2,0.1806720018386841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,1,power_law_1.2,0.10826879739761353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,1,power_law_1.2,0.1925503969192505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,1,power_law_1.2,0.11730560064315795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,1,power_law_1.2,0.1903488039970398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,1,power_law_1.2,0.11758079528808593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,1,power_law_1.2,0.17782399654388428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,1,power_law_1.2,0.11258879899978638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,1,power_law_1.2,0.18083200454711915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,1,power_law_1.2,0.11101440191268921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,1,power_law_1.2,0.18686720132827758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,1,power_law_1.2,0.11311999559402466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,1,power_law_1.2,0.1778432011604309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,1,power_law_1.2,0.11367679834365844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,1,power_law_1.2,0.19020800590515136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,1,power_law_1.2,0.12006399631500245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,1,power_law_1.2,0.20714879035949707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,1,power_law_1.2,0.12904319763183594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,1,power_law_1.2,0.2081216096878052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,1,power_law_1.2,0.1410752058029175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,1,power_law_1.2,0.2168639898300171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,1,power_law_1.2,0.142086398601532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,1,power_law_1.2,0.2369920015335083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,1,power_law_1.2,0.15302400588989257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,1,power_law_1.2,0.2727679967880249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,1,power_law_1.2,0.1727231979370117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,1,power_law_1.2,0.3251967906951904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,1,power_law_1.2,0.20236799716949463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,1,power_law_1.2,0.3874943971633911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,1,power_law_1.2,0.24959359169006348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,1,power_law_1.2,0.4451007843017578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,1,power_law_1.2,0.2850944042205811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,1,power_law_1.2,0.057683199644088745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,1,power_law_1.2,0.6049280166625977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,1,power_law_1.2,0.3837183952331543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,1,power_law_1.2,0.02077440023422241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,1,power_law_1.2,0.06617599725723267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,1,power_law_1.2,0.769977617263794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,1,power_law_1.2,0.4791872024536133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,1,power_law_1.2,0.022681599855422972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,1,power_law_1.2,0.08419839739799499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,1,power_law_1.2,1.035424041748047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,1,power_law_1.2,0.6667903900146485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,1,power_law_1.2,0.021798400580883025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,1,power_law_1.2,0.10894080400466918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,1,power_law_1.2,1.3395263671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,1,power_law_1.2,0.8515071868896484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,1,power_law_1.2,0.02316800057888031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,1,power_law_1.2,0.11306240558624267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,1,power_law_1.2,1.897817611694336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,1,power_law_1.2,1.2380800247192383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,1,power_law_1.2,0.026092800498008727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,1,power_law_1.2,0.11764479875564575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,1,power_law_1.2,1.597100830078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,1,power_law_1.2,2.489279937744141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,1,power_law_1.2,0.03049600124359131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,1,power_law_1.2,0.11581439971923828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,1,power_law_1.2,0.028972798585891725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,1,power_law_1.2,2.001900863647461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,1,power_law_1.2,3.0571136474609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,1,power_law_1.2,0.12023040056228637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,1,power_law_1.2,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,1,power_law_1.2,3.1158143997192385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,1,power_law_1.2,0.1134719967842102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,1,power_law_1.2,4.837299346923828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,1,power_law_1.2,0.02985599935054779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,1,power_law_1.2,0.1183743953704834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,1,power_law_1.2,6.150969696044922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,1,power_law_1.2,0.03052160143852234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,1,power_law_1.2,9.686822509765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,1,power_law_1.2,0.11498880386352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,1,power_law_1.2,0.033670398592948916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,1,power_law_1.2,0.11694079637527466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,1,power_law_1.2,0.032358399033546446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,1,power_law_1.2,0.11628799438476563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,1,power_law_1.2,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,1,power_law_1.2,0.12641919851303102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,1,power_law_1.2,0.03685120046138764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,1,power_law_1.2,0.13452800512313842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,1,power_law_1.2,0.06157439947128296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,1,power_law_1.2,0.14213119745254515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,1,power_law_1.2,0.07625600099563598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,1,power_law_1.2,0.14788479804992677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,1,power_law_1.2,0.07920640110969543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,1,power_law_1.2,0.16915839910507202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,1,power_law_1.2,0.05877760052680969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.2,0.1916416049003601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.2,0.06060799956321716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.2,0.22889599800109864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.2,0.08561919927597046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.2,0.2799551963806152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.2,0.10030720233917237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.2,0.3603775978088379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.2,0.13671040534973145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.2,0.4388160228729248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.2,0.16851840019226075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.2,0.5989759922027588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.2,0.24377601146697997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.2,0.7629439830780029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.2,0.3148159980773926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.2,1.1140416145324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.2,0.45372800827026366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.2,1.43953275680542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.2,0.5941504001617431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.2,1.7694784164428712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.2,0.7342463970184326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,1,power_law_1.2,2.7855615615844727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,1,power_law_1.2,1.1570560455322265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,1,power_law_1.2,2.270438385009766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,1,power_law_1.2,5.495321655273438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,4,balanced,0.21465599536895752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,4,balanced,0.2411093314488729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,4,balanced,0.1495306690533956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,4,balanced,0.34276266892751056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,4,balanced,0.15068266789118448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,4,balanced,0.33984533945719403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,4,balanced,0.21057599782943726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,4,balanced,0.3280106584231059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,4,balanced,0.1989333430926005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,4,balanced,0.35656531651814777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,4,balanced,0.1896160046259562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,4,balanced,0.3739519913991292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,4,balanced,0.18939733505249023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,4,balanced,0.31458133459091187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,4,balanced,0.18900267283121744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,4,balanced,0.33450134595235187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,4,balanced,0.19829332828521729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,4,balanced,0.3102239966392517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,4,balanced,0.19595734278361002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,4,balanced,0.2959573268890381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,4,balanced,0.1880319913228353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,4,balanced,0.2969813346862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,4,balanced,0.18156800667444864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,4,balanced,0.3033333420753479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,4,balanced,0.19097065925598145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,4,balanced,0.2741173307100932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,4,balanced,0.2674506704012553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,4,balanced,0.1719520092010498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,4,balanced,0.2609440088272095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,4,balanced,0.1809013287226359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,4,balanced,0.2589866717656453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,4,balanced,0.169429341952006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,4,balanced,0.4771626790364583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,4,balanced,0.1622933348019918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,4,balanced,0.16474133729934692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,4,balanced,0.43450133005777997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,4,balanced,0.2898079951604207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,4,balanced,0.5724853277206421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,4,balanced,0.269813338915507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,4,balanced,0.04764266808827718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,4,balanced,0.7602933247884115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,4,balanced,0.13126933574676514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,4,balanced,0.048325334986050926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,4,balanced,0.11621866623560588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,4,balanced,0.3479306697845459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,4,balanced,0.04752000172932943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,4,balanced,1.0762720108032227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,4,balanced,0.1913599967956543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,4,balanced,0.05816000203291575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,4,balanced,0.4532159964243571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,4,balanced,0.18221867084503174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,4,balanced,0.058821335434913635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,4,balanced,0.18125865856806436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,4,balanced,1.426751931508382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,4,balanced,0.059088001648585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,4,balanced,0.6233546733856201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,4,balanced,0.19445333878199259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,4,balanced,0.07672533392906189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,4,balanced,0.1932906707127889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,4,balanced,0.06689066688219707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,4,balanced,2.088101387023926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,4,balanced,0.8071093559265137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,4,balanced,0.20433066288630167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,4,balanced,0.1323040028413137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,4,balanced,0.18890132506688437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,4,balanced,0.13365333278973898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,4,balanced,0.19208532571792603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,4,balanced,2.754309336344401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,4,balanced,1.1575626532236736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,4,balanced,0.13502400120099387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,4,balanced,0.1784693400065104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,4,balanced,0.13635733723640442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,4,balanced,0.1802826722462972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,4,balanced,0.127920001745224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,4,balanced,0.18378132581710815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,4,balanced,1.48688538869222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,4,balanced,0.10048533479372661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,4,balanced,4.09773317972819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,4,balanced,0.1585813363393148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,4,balanced,0.14519466956456503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,4,balanced,0.15584533413251242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,4,balanced,0.12213333447774251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,4,balanced,0.14273066322008768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,4,balanced,0.1597599983215332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,4,balanced,0.13243732849756876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,4,balanced,2.1140054066975913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,4,balanced,0.20765332380930582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,4,balanced,0.2288693388303121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,4,balanced,0.2665226658185323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,4,balanced,0.19075733423233032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,4,balanced,0.375434676806132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,4,balanced,0.25650133689244586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,4,balanced,0.49435198307037354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,4,balanced,0.3251306613286336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,4,balanced,2.6856746673583984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,4,balanced,0.7054933706919352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,4,balanced,0.4442880153656006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,4,balanced,0.9303840001424154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,4,balanced,0.559989333152771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,4,balanced,1.3680960337320964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,4,balanced,0.800538698832194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,4,balanced,3.269551912943522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,4,balanced,1.041167974472046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,4,balanced,1.8079946835835774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,4,balanced,2.6735251744588218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,4,balanced,1.5129493077596028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,4,balanced,4.938058535257976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,4,balanced,3.5492852528889975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,4,balanced,1.9746559460957844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,4,balanced,4.420442581176758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,4,balanced,2.4563306172688804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,4,balanced,7.038096110026042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,4,balanced,3.860549290974935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,4,balanced,9.27571169535319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,4,balanced,14.023061116536459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,4,balanced,7.60264523824056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,4,power_law_1.01,0.301311993598938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,4,power_law_1.01,0.2560895919799805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,4,power_law_1.01,0.2813568115234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,4,power_law_1.01,0.3459968090057373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,4,power_law_1.01,0.15912959575653077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,4,power_law_1.01,0.16312320232391359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,4,power_law_1.01,0.17662719488143921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,4,power_law_1.01,0.16952320337295532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,4,power_law_1.01,0.33026559352874757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,4,power_law_1.01,0.15931520462036133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,4,power_law_1.01,0.1706431984901428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,4,power_law_1.01,0.3254143953323364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,4,power_law_1.01,0.19139839410781861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,4,power_law_1.01,0.19752320051193237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,4,power_law_1.01,0.3177472114562988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,4,power_law_1.01,0.19680000543594361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,4,power_law_1.01,0.18770560026168823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,4,power_law_1.01,0.327513599395752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,4,power_law_1.01,0.1899135947227478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,4,power_law_1.01,0.3067392110824585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,4,power_law_1.01,0.19413119554519653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,4,power_law_1.01,0.30213758945465086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,4,power_law_1.01,0.18504960536956788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,4,power_law_1.01,0.300383996963501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,4,power_law_1.01,0.1843392014503479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,4,power_law_1.01,0.177401602268219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,4,power_law_1.01,0.34182400703430177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,4,balanced,5.358394622802734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,4,power_law_1.01,0.2063999891281128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,4,power_law_1.01,0.32826240062713624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,4,power_law_1.01,0.19141119718551636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,4,power_law_1.01,0.37651839256286623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,4,power_law_1.01,0.20288639068603515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,4,power_law_1.01,0.24349439144134521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,4,power_law_1.01,0.490342378616333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,4,power_law_1.01,0.28882560729980467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,4,power_law_1.01,0.44202241897583006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,4,power_law_1.01,0.3204096078872681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,4,balanced,6.647189458211263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,4,power_law_1.01,0.5403200149536133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,4,power_law_1.01,0.3913280010223389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,4,power_law_1.01,0.7330048084259033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,4,power_law_1.01,0.5531007766723632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,4,power_law_1.01,0.7344704151153565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,4,power_law_1.01,1.0220288276672362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,4,power_law_1.01,0.8599871635437012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,4,power_law_1.01,1.2529472351074218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,4,power_law_1.01,1.20382719039917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,4,power_law_1.01,1.7124095916748048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,4,power_law_1.01,1.5392191886901856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,4,balanced,10.247386932373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,4,power_law_1.01,2.262873649597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,4,power_law_1.01,2.6251903533935548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,4,power_law_1.01,2.9048063278198244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,4,power_law_1.01,4.299987030029297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,4,power_law_1.01,4.152025604248047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,4,power_law_1.01,4.998624038696289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,4,power_law_1.01,6.0928703308105465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,4,power_law_1.01,6.2657215118408205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,4,power_law_1.01,9.320121765136719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,4,power_law_1.01,11.106246185302734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,4,power_law_1.01,9.709113311767577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,4,power_law_1.01,13.119187927246093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,4,power_law_1.01,25.565113830566407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,4,power_law_1.01,13.962200927734376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,4,balanced,18.23046366373698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,4,power_law_1.01,27.613153076171876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,4,power_law_1.01,40.49903259277344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,4,power_law_1.2,0.22401280403137208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,4,power_law_1.01,0.0456959992647171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,4,power_law_1.2,0.30474879741668703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,4,power_law_1.01,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,4,power_law_1.2,0.35047039985656736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,4,power_law_1.01,0.045612800121307376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,4,power_law_1.2,0.3318079948425293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,4,power_law_1.01,0.05550079941749573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,4,power_law_1.2,0.33766400814056396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,4,power_law_1.01,0.05735039710998535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,4,power_law_1.2,0.3381632089614868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,4,power_law_1.01,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,4,power_law_1.2,0.31658239364624025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,4,power_law_1.01,0.07505279779434204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,4,power_law_1.2,0.30838398933410643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,4,power_law_1.01,0.06664320230484008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,4,power_law_1.2,0.18112640380859374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,4,power_law_1.2,0.30943999290466306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,4,power_law_1.01,0.1295040011405945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,4,power_law_1.2,0.18309119939804078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,4,power_law_1.2,0.31059839725494387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,4,power_law_1.01,0.13044480085372925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,4,power_law_1.2,0.17443840503692626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,4,power_law_1.2,0.28229761123657227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,4,power_law_1.01,0.13518719673156737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,4,power_law_1.2,0.21417601108551027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,4,power_law_1.2,0.31468799114227297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,4,power_law_1.01,0.141593599319458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,4,power_law_1.2,0.19066879749298096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,4,power_law_1.2,0.34778881072998047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,4,power_law_1.01,0.12993919849395752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,4,power_law_1.2,0.19454079866409302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,4,power_law_1.2,0.3539776086807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,4,power_law_1.01,0.09992319941520691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,4,power_law_1.2,0.19243520498275757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,4,power_law_1.2,0.5115007877349853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,4,power_law_1.01,0.14711040258407593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,4,power_law_1.2,0.1901952028274536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,4,power_law_1.2,0.46499199867248536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,4,power_law_1.01,0.12269439697265624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,4,power_law_1.2,0.18674559593200685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,4,power_law_1.2,0.5637184143066406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,4,power_law_1.01,0.1593727946281433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,4,power_law_1.2,0.18465280532836914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,4,power_law_1.01,0.1914240002632141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,4,power_law_1.2,0.663801622390747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,4,power_law_1.01,0.20543360710144043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,4,power_law_1.2,0.17149440050125123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,4,power_law_1.01,0.19296640157699585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,4,power_law_1.2,0.7254079818725586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.01,0.2665152072906494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,4,power_law_1.2,0.1874943971633911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,4,power_law_1.01,0.18279039859771729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,4,power_law_1.2,1.5529919624328614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.01,0.3764224052429199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,4,power_law_1.2,0.19445120096206664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,4,power_law_1.01,0.18110719919204712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,4,power_law_1.2,1.9677759170532227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.01,0.4920447826385498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,4,power_law_1.2,0.2028287887573242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,4,power_law_1.01,0.17715200185775756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,4,power_law_1.2,2.950764846801758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.01,0.7076352119445801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,4,power_law_1.2,0.23887999057769777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,4,power_law_1.01,0.16947200298309326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.01,0.9272383689880371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,4,power_law_1.2,3.9610111236572267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,4,power_law_1.01,0.162336003780365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,4,power_law_1.2,0.2810944080352783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,4,power_law_1.01,0.18290560245513915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.01,1.3643775939941407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,4,power_law_1.2,0.3510272026062012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,4,power_law_1.2,4.588032150268555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,4,power_law_1.01,0.20556159019470216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,4,power_law_1.2,0.4648575782775879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.01,1.8016576766967773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,4,power_law_1.2,6.117427062988281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,4,power_law_1.01,0.20297598838806152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,4,power_law_1.2,0.5943295955657959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.01,2.675699234008789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,4,power_law_1.01,0.2587712049484253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,4,power_law_1.2,0.6680255889892578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,4,power_law_1.2,10.1517822265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.01,3.553094482421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,4,power_law_1.01,0.23444480895996095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,4,power_law_1.2,0.9608448028564454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,4,power_law_1.01,4.422700881958008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,4,power_law_1.01,0.23543040752410888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,4,power_law_1.2,15.231840515136719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,4,power_law_1.2,1.4867712020874024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,4,power_law_1.01,0.3346240043640137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,4,power_law_1.01,7.049523162841797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,4,power_law_1.2,1.5477375984191895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.01,0.36620800495147704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,4,power_law_1.2,18.71769561767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,4,power_law_1.2,2.5023359298706054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.01,0.5467904090881348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,4,power_law_1.01,14.044160461425781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,4,power_law_1.2,3.709107208251953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.01,0.6021056175231934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,4,power_law_1.2,28.711410522460938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.01,0.9293631553649903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,4,power_law_1.2,4.661868667602539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.01,1.206758403778076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,4,power_law_1.2,6.3589118957519535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.01,1.8772159576416017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,4,power_law_1.2,7.709779357910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.01,2.375315284729004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,4,power_law_1.2,51.63386840820313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.01,3.915974426269531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,4,power_law_1.2,14.46776885986328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.01,4.779014587402344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,4,power_law_1.01,7.0823616027832035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,4,power_law_1.2,28.886502075195313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,4,power_law_1.01,8.632978820800782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,4,power_law_1.01,20.137113952636717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,4,power_law_1.2,0.21368319988250734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,4,power_law_1.2,0.1739776015281677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,4,power_law_1.2,0.15166720151901245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,4,power_law_1.2,0.19754879474639891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,4,power_law_1.2,0.20026240348815919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,4,power_law_1.2,0.1864127993583679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,4,power_law_1.2,0.045075199007987975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,4,power_law_1.2,0.1814911961555481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,4,power_law_1.2,0.184716796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,4,power_law_1.2,0.045817598700523376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,4,power_law_1.2,0.17804800271987914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,4,power_law_1.2,0.044684800505638125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,4,power_law_1.2,0.15827200412750245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,4,power_law_1.2,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,4,power_law_1.2,0.15482879877090455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,4,power_law_1.2,0.05740159749984741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,4,power_law_1.2,0.1978943943977356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,4,power_law_1.2,0.058771198987960814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,4,power_law_1.2,0.20894079208374022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,4,power_law_1.2,0.07482240200042725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,2,balanced,0.13913066188494363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,4,power_law_1.2,0.19657599925994873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,4,power_law_1.2,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,2,balanced,0.18028799692789713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,2,balanced,0.08596799770991008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,4,power_law_1.2,0.24023680686950682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,2,balanced,0.2953333258628845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,4,power_law_1.2,0.12973439693450928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,2,balanced,0.11757333079973857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,4,power_law_1.2,0.2560767889022827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,2,balanced,0.3040906588236491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,4,power_law_1.2,0.13027199506759643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,2,balanced,0.2023680011431376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,4,power_law_1.2,0.25994880199432374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,2,balanced,0.3237386743227641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,2,balanced,0.17406932512919107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,4,power_law_1.2,0.13416320085525513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,4,power_law_1.2,0.31957120895385743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,2,balanced,0.30447999636332196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,2,balanced,0.18420799573262533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,4,power_law_1.2,0.13763200044631957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,2,balanced,0.3190079927444458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.2,0.4084159851074219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,2,balanced,0.18073066075642905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,4,power_law_1.2,0.13092479705810547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,2,balanced,0.29629333813985187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.2,0.6488448143005371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,2,balanced,0.1841599941253662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,4,power_law_1.2,0.10045440196990967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,2,balanced,0.297706663608551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,2,balanced,0.1796320080757141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.2,0.9309056282043457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,4,power_law_1.2,0.1446336030960083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,2,balanced,0.3015519976615906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,2,balanced,0.19401599963506064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.2,1.1752063751220703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,4,power_law_1.2,0.1223680019378662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,2,balanced,0.3006719946861267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,2,balanced,0.1774133245150248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.2,1.35928316116333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,4,power_law_1.2,0.1581120014190674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,2,balanced,0.2935199936230977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,2,balanced,0.1721173326174418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.2,2.3067712783813477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,4,power_law_1.2,0.2057408094406128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,2,balanced,0.2958559989929199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,2,balanced,0.17098132769266763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.2,3.1668800354003905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.2,0.2653568029403687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,2,balanced,0.2666719953219096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,2,balanced,0.16960533459981283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.2,4.189542388916015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,2,balanced,0.2621440092722575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,2,balanced,0.16265599926312765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,2,balanced,0.261626660823822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,2,balanced,0.1649386684099833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.2,6.1955711364746096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,2,balanced,0.26181866725285846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,2,balanced,0.1637386679649353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,4,power_law_1.2,6.0464832305908205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,2,balanced,0.1694613297780355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,2,balanced,0.4116640090942383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,4,power_law_1.2,9.917510223388671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,2,balanced,0.2739359935124715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,2,balanced,0.40068264802296955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,2,balanced,0.2469279964764913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,2,balanced,0.5448053280512491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,4,power_law_1.2,20.488397216796876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,2,balanced,0.3176213304201762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,2,balanced,0.7198879718780518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,2,balanced,0.39879465103149414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,2,balanced,1.0073440074920654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,2,balanced,0.5322719812393188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,2,balanced,1.3051893711090088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,2,balanced,0.6862293084462484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,2,balanced,1.8672213554382324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,2,balanced,0.9626239935557047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,2,balanced,2.37062406539917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,2,balanced,1.2459093729654949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,2,balanced,3.335280100504557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,2,balanced,1.8099573453267415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,2,balanced,4.263845443725586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,2,balanced,2.3666507403055825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,2,balanced,5.16706657409668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,2,balanced,2.9045867919921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,2,balanced,7.954986572265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,2,balanced,4.62116273244222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,2,balanced,15.822303771972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,2,balanced,9.11900774637858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,2,balanced,0.083514670530955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,2,balanced,0.11768533786137898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,2,balanced,0.1975253423055013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,2,balanced,0.16802134116490683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,2,balanced,0.16244799892107645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,2,balanced,0.16396799683570862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,2,balanced,0.1885546644528707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,2,balanced,0.18964266777038574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,2,balanced,0.18745599190394083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,2,balanced,0.1737920045852661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,2,balanced,0.1683466633160909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,2,balanced,0.17068799336751303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,2,balanced,0.16330132881800333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,2,balanced,0.1574133336544037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,2,balanced,0.1439626713593801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,2,balanced,0.14192000031471252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,2,balanced,0.12820800145467123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,2,balanced,0.24277333418528238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,2,balanced,0.18985066811243692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,2,balanced,0.24221332867940268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,2,balanced,0.30248000224431354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,2,balanced,0.42374932765960693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,2,balanced,0.5369600057601929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,2,balanced,0.7774346669514974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,2,balanced,1.0143520037333171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,2,balanced,1.4939306577046711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,2,balanced,1.982485294342041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,2,balanced,0.028783999383449554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,2,power_law_1.01,0.16906239986419677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.2,0.3743232011795044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,2,balanced,0.030026666820049286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,2,power_law_1.01,0.19875839948654175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.2,0.48825597763061523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,2,balanced,0.02924799919128418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,2,power_law_1.01,0.24412159919738768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,2,balanced,2.449903964996338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.2,0.704204797744751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,2,balanced,0.03523733218510946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,2,power_law_1.01,0.31935360431671145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.2,0.9286848068237304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,2,balanced,0.046581332882245384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,2,power_law_1.01,0.32233600616455077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.2,1.3667712211608887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,2,balanced,0.04584000011285146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,2,power_law_1.01,0.3107903957366943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,2,balanced,0.04085866610209147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.2,1.8010175704956055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,2,power_law_1.01,0.3052671909332275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,2,balanced,0.04814399778842926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,2,power_law_1.01,0.2914367914199829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.2,2.6798847198486326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,2,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,2,power_law_1.01,0.29788799285888673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.2,3.5531455993652346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,2,balanced,0.046853333711624146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,2,power_law_1.01,0.3071743965148926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,4,power_law_1.2,4.434732818603516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,2,balanced,0.05605866511662801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,2,power_law_1.01,0.2849024057388306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,2,balanced,0.10378666718800862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,2,balanced,3.8577121098836265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,2,power_law_1.01,0.31402881145477296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,4,power_law_1.2,7.0572669982910154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,2,balanced,0.10539733370145161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,2,power_law_1.01,0.3092736005783081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,2,balanced,0.13340800007184347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,2,power_law_1.01,0.33357439041137693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,4,power_law_1.2,14.059730529785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,2,balanced,0.09391466776529948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,2,power_law_1.01,0.35980160236358644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,2,power_law_1.01,0.3783679962158203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,2,balanced,0.07558399935563405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,2,power_law_1.01,0.4463359832763672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,2,balanced,0.09449600179990132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,2,power_law_1.01,0.5609600067138671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,2,balanced,0.11540266871452332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,2,power_law_1.01,0.674131202697754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,2,balanced,0.14103999733924866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,2,power_law_1.01,0.8010239601135254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,2,balanced,0.19473065932591757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,2,power_law_1.01,1.1842816352844239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,2,power_law_1.01,1.6607679367065429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,2,balanced,0.24682132403055826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,2,power_law_1.01,1.609343910217285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,2,balanced,0.345909317334493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,2,power_law_1.01,2.353702354431152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,2,power_law_1.01,3.1148351669311523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,2,balanced,0.44124801953633624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,2,power_law_1.01,4.941004943847656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,2,balanced,7.725605646769206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,2,balanced,0.6422079801559448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,2,power_law_1.01,5.716287994384766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,2,power_law_1.01,7.744364929199219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,2,balanced,0.8433866500854492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,2,power_law_1.01,12.047206115722656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,2,balanced,1.2356213728586833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,2,power_law_1.01,25.697048950195313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,2,balanced,1.622927983601888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,2,balanced,2.0138293902079263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,2,balanced,3.2076212565104165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,2,balanced,6.349744160970052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,2,power_law_1.01,0.10124800205230713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,2,power_law_1.01,0.10915199518203736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,2,power_law_1.01,0.15115519762039184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,2,power_law_1.01,0.18394240140914916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,2,power_law_1.01,0.1824447989463806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,2,power_law_1.01,0.18805760145187378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,2,power_law_1.01,0.18581759929656982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,2,power_law_1.01,0.11197439432144166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,2,power_law_1.01,0.1836159944534302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,2,power_law_1.01,0.11190400123596192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,2,power_law_1.01,0.1818943977355957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,2,power_law_1.01,0.11346559524536133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,2,power_law_1.01,0.17420159578323363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,2,power_law_1.01,0.1888576030731201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,2,power_law_1.01,0.18042880296707153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,2,power_law_1.01,0.18057600259780884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,2,power_law_1.01,0.18542079925537108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,2,power_law_1.01,0.1828160047531128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,2,power_law_1.01,0.19982080459594725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,2,power_law_1.01,0.026259198784828186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,2,power_law_1.01,0.17404799461364745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,2,power_law_1.01,0.19474560022354126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,2,power_law_1.01,0.027027198672294618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,2,power_law_1.01,0.18021119832992555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,2,power_law_1.01,0.23512959480285645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,2,power_law_1.01,0.026387199759483337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,2,power_law_1.01,0.1778175950050354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,2,power_law_1.01,0.24102399349212647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,2,power_law_1.01,0.031219199299812317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,2,power_law_1.01,0.1735360026359558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,2,power_law_1.01,0.2734783887863159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,2,power_law_1.01,0.044326400756835936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,2,power_law_1.01,0.16531840562820435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,2,power_law_1.01,0.32126080989837646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,2,power_law_1.01,0.04296320080757141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,2,power_law_1.01,0.17543679475784302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,2,power_law_1.01,0.4026175975799561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,2,power_law_1.01,0.03923200070858002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,2,power_law_1.01,0.18365440368652344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,2,power_law_1.01,0.48801279067993164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,2,power_law_1.01,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,2,power_law_1.01,0.19029120206832886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,2,power_law_1.01,0.5519360065460205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,2,power_law_1.01,0.052985602617263795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,2,power_law_1.01,0.20342400074005126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,2,power_law_1.01,0.6805952072143555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,2,power_law_1.01,0.04421760141849518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,2,power_law_1.01,0.2131200075149536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,2,power_law_1.01,0.05397760272026062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,2,power_law_1.01,0.9901887893676757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,2,power_law_1.01,0.2536832094192505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,2,power_law_1.01,0.10408960580825806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,2,power_law_1.01,0.2556864023208618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,2,power_law_1.01,1.512992000579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,2,power_law_1.01,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.01,0.3232896089553833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,2,power_law_1.01,1.8901504516601562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,2,power_law_1.01,0.13317760229110717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.01,0.37813758850097656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,2,power_law_1.01,2.7964223861694335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,2,power_law_1.01,0.09175040125846863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.01,0.4780928134918213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,2,power_law_1.01,3.4988094329833985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,2,power_law_1.01,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.01,0.750489616394043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,2,power_law_1.01,0.09280639886856079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,2,power_law_1.01,4.2337791442871096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.01,0.7291647911071777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,2,power_law_1.01,0.1146496057510376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.01,1.2334848403930665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,2,power_law_1.01,5.700614547729492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.01,0.14094719886779786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.01,1.7686271667480469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.01,0.19489279985427857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.01,2.2702783584594726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,2,power_law_1.01,14.769062805175782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.01,0.24601600170135499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.01,3.127987289428711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.01,0.3480832099914551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,2,power_law_1.01,3.7523006439208983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.01,0.44347519874572755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.01,0.6425727844238281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,2,power_law_1.01,5.981036758422851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.01,0.8376768112182618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.01,1.233683204650879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,2,power_law_1.01,12.98151092529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.01,1.6262336730957032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,2,power_law_1.01,2.016870307922363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,2,power_law_1.01,3.1978559494018555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,2,power_law_1.01,6.335750579833984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,2,power_law_1.2,0.17807999849319459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,2,power_law_1.2,0.19582079648971557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,2,power_law_1.2,0.22685439586639405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,2,power_law_1.2,0.319980788230896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,2,power_law_1.2,0.3245376110076904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,2,power_law_1.2,0.3203455924987793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,2,power_law_1.2,0.3105792045593262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,2,power_law_1.2,0.3064768075942993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,2,power_law_1.2,0.3070208072662354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,2,power_law_1.2,0.3085184097290039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,2,power_law_1.2,0.2915712118148804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,2,power_law_1.2,0.30935680866241455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,2,power_law_1.2,0.3270080089569092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,2,power_law_1.2,0.33016960620880126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,2,power_law_1.2,0.3652992010116577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,2,power_law_1.2,0.38428800106048583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,2,power_law_1.2,0.45022082328796387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,2,power_law_1.2,0.5951807975769043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,2,power_law_1.2,0.6837503910064697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,2,power_law_1.2,0.930784034729004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,2,power_law_1.2,1.1166080474853515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,2,power_law_1.2,1.6276607513427734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,2,power_law_1.2,2.269785690307617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,2,power_law_1.2,3.0174591064453127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,2,power_law_1.2,3.5173313140869142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,2,power_law_1.2,5.304147338867187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,2,power_law_1.2,7.047596740722656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,2,power_law_1.2,8.142745971679688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,2,power_law_1.2,13.260665893554688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,2,power_law_1.2,24.97034912109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,2,power_law_1.2,0.10002559423446655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,2,power_law_1.2,0.10451840162277222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,2,power_law_1.2,0.11661440134048462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,2,power_law_1.2,0.10083199739456176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,2,power_law_1.2,0.14822399616241455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,2,power_law_1.2,0.15765759944915772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,2,power_law_1.2,0.19007359743118285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,2,power_law_1.2,0.1783360004425049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,2,power_law_1.2,0.1802559971809387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,2,power_law_1.2,0.17303040027618408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,2,power_law_1.2,0.19173760414123536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,2,power_law_1.2,0.17668479681015015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,2,power_law_1.2,0.18045439720153808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,2,power_law_1.2,0.17766400575637817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,2,power_law_1.2,0.18417279720306395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,2,power_law_1.2,0.026995199918746948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,2,power_law_1.2,0.17653119564056396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,2,power_law_1.2,0.18392959833145142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,2,power_law_1.2,0.028288000822067262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,2,power_law_1.2,0.16722559928894043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,2,power_law_1.2,0.17948800325393677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,2,power_law_1.2,0.026918399333953857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,2,power_law_1.2,0.15920000076293944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,2,power_law_1.2,0.17950719594955444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,2,power_law_1.2,0.03190400004386902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,2,power_law_1.2,0.18437119722366332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,2,power_law_1.2,0.17764480113983155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,2,power_law_1.2,0.04245760142803192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,2,power_law_1.2,0.18503040075302124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,2,power_law_1.2,0.04437119960784912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,2,power_law_1.2,0.20424959659576417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,2,power_law_1.2,0.038950398564338684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,2,power_law_1.2,0.2
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,2,power_law_1.2,0.04578559994697571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,2,power_law_1.2,0.2368000030517578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,2,power_law_1.2,0.05025280117988586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,2,power_law_1.2,0.23697280883789062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,2,power_law_1.2,0.04410240054130554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,2,power_law_1.2,0.26218879222869873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,2,power_law_1.2,0.05192959904670715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,2,power_law_1.2,0.32839040756225585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,2,power_law_1.2,0.10289280414581299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,2,power_law_1.2,0.3789184093475342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,2,power_law_1.2,0.10325759649276733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,2,power_law_1.2,0.48099842071533205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,2,power_law_1.2,0.13222399950027466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,2,power_law_1.2,0.582047986984253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,2,power_law_1.2,0.09134079813957215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,2,power_law_1.2,0.8202048301696777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,2,power_law_1.2,0.0737600028514862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,2,power_law_1.2,1.0774208068847657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,2,power_law_1.2,0.09400960206985473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,2,power_law_1.2,1.552614402770996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,2,power_law_1.2,0.11557120084762573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,2,power_law_1.2,1.989344024658203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.2,0.13929599523544312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,2,power_law_1.2,3.157683181762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.2,0.19575040340423583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,2,power_law_1.2,3.280108642578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.2,0.24679040908813477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,2,power_law_1.2,4.35098876953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.2,0.34581120014190675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.2,0.4419839859008789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,2,power_law_1.2,8.297695922851563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.2,0.6428607940673828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.2,0.8401215553283692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,2,power_law_1.2,13.3157958984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.2,1.235372829437256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.2,1.6266624450683593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,2,power_law_1.2,2.0188224792480467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,2,power_law_1.2,3.193996810913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,2,power_law_1.2,6.3348033905029295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,1,balanced,0.10195733110109965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,1,balanced,0.17482133706410727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,1,balanced,0.2844533324241638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,1,balanced,0.2828373312950134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,1,balanced,0.32393600543340045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,1,balanced,0.2871200044949849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,1,balanced,0.3011893431345622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,1,balanced,0.2874026695887248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,1,balanced,0.3041759928067525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,1,balanced,0.28622933228810626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,1,balanced,0.2827039957046509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,1,balanced,0.2730453411738078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,1,balanced,0.26629867156346637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,1,balanced,0.2590186595916748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,1,balanced,0.25598933299382526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,1,balanced,0.25439999500910443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,1,balanced,0.2579519947369893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,1,balanced,0.41486934820810956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,1,balanced,0.37057065963745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,1,balanced,0.49649067719777423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,1,balanced,0.6254293521245321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,1,balanced,0.857754627863566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,1,balanced,1.0962026913960774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,1,balanced,1.5770293871561687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,1,balanced,2.066714604695638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,1,balanced,3.024943987528483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,1,balanced,3.997722625732422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,1,balanced,4.919530550638835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,1,balanced,0.08029866715272267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,1,balanced,0.11657599608103435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,1,balanced,7.876837412516276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,1,balanced,0.16209600369135538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,1,balanced,0.16526400049527487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,1,balanced,0.1668000022570292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,1,balanced,0.1824000080426534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,1,balanced,0.18200532595316568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,1,balanced,0.08318933347860973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,1,balanced,0.17429866393407187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,1,balanced,0.10783466696739197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,1,balanced,0.18737600247065225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,1,balanced,0.19449067115783691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,1,balanced,0.1702559987703959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,1,balanced,0.16686399777730307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,1,balanced,0.17336533466974893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,1,balanced,0.1572480003039042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,1,balanced,0.1572426656881968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,1,balanced,0.1764799952507019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,1,balanced,0.16158933440844217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,1,balanced,0.1704960068066915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,1,balanced,0.17674134174982706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,1,balanced,0.1578879952430725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,1,balanced,0.157642662525177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,1,balanced,0.1600160002708435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,1,balanced,0.17076265811920166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,1,balanced,0.16158933440844217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,1,balanced,0.17404266198476157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,1,balanced,0.1695786714553833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,1,balanced,0.15779733657836914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,1,balanced,0.250271995862325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,1,balanced,15.938986460367838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,1,balanced,0.15067733327547708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,1,balanced,0.23106666405995688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,1,balanced,0.1492800017197927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,1,balanced,0.13902933398882547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,1,balanced,0.2960053284962972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,1,balanced,0.14061333735783896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,1,balanced,0.370037317276001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,1,balanced,0.13893333077430725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,1,balanced,0.23540266354878744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,1,balanced,0.5112853447596232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,1,balanced,0.199562668800354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,1,balanced,0.653493324915568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,1,balanced,0.2627573410669963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,2,power_law_1.2,0.19569920301437377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,1,balanced,0.32256533702214557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,1,balanced,0.9585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,2,power_law_1.2,0.19081599712371827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,1,balanced,0.4495946566263835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,2,power_law_1.2,0.21119999885559082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,2,power_law_1.2,0.21132159233093262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,1,balanced,1.2401173114776611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,1,balanced,0.5662560065587362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,2,power_law_1.2,0.22636799812316893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,2,power_law_1.2,0.28923521041870115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,1,balanced,0.8242239952087402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.2,0.3154623985290527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,1,balanced,1.8382666905721028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.2,0.3975167989730835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.2,0.5256192207336425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,1,balanced,1.0783092975616455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.2,0.7352831840515137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.2,0.7929088115692139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,1,balanced,2.420479933420817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.2,1.1773504257202148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.2,1.5993791580200196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,1,balanced,1.5920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.2,2.5625471115112304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.2,3.21130256652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,1,balanced,2.9999094009399414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,2,power_law_1.2,4.078265762329101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,2,power_law_1.2,6.071136093139648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,1,balanced,2.1201440493265786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,2,power_law_1.2,10.895584106445312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,1,balanced,4.732927958170573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,1,balanced,2.6153386433919272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,1,balanced,4.176202774047852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,1,balanced,9.459802627563477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,1,balanced,8.481130599975586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,1,balanced,0.023226665953795116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,1,balanced,0.02603200078010559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,1,balanced,0.02629866699377696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,1,balanced,0.02664000044266383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,1,balanced,0.03315199911594391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,1,balanced,0.04109866668780645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,1,balanced,0.03615466753641764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,1,balanced,0.04075733323891958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,1,balanced,0.036746665835380554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,1,balanced,0.03852266569932302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,1,balanced,0.044591998060544334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,1,balanced,0.042223999897638954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,1,balanced,0.042853335539499916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,1,balanced,0.04859200119972229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,1,balanced,0.10179733236630757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,1,balanced,0.10416533549626668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,1,balanced,0.12922133008639017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,1,balanced,0.07406400144100189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,1,balanced,0.09154133001963298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,1,balanced,0.12448533376057942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,1,balanced,0.14638400077819824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,1,balanced,0.20484266678492227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,1,balanced,0.2574613293011983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,1,power_law_1.01,0.1009984016418457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,1,balanced,0.36715734004974365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,1,power_law_1.01,0.13903360366821288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,1,power_law_1.01,0.1850111961364746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,1,balanced,0.47544535001118976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,1,power_law_1.01,0.2749119997024536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,1,power_law_1.01,0.3037823915481567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,1,power_law_1.01,0.2821439981460571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,1,power_law_1.01,0.30108160972595216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,1,balanced,0.6936266422271729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,1,power_law_1.01,0.2956032037734985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,1,power_law_1.01,0.2961344003677368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,1,power_law_1.01,0.28155519962310793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,1,power_law_1.01,0.2823231935501099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,1,balanced,0.9131306807200114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,1,power_law_1.01,0.2863744020462036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,1,power_law_1.01,0.27083520889282225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,1,power_law_1.01,0.31976320743560793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,1,power_law_1.01,0.32953600883483886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,1,power_law_1.01,0.3396735906600952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,1,power_law_1.01,0.07948799729347229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,1,power_law_1.01,0.39616639614105226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,1,balanced,1.129050652186076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,1,power_law_1.01,0.08367999792098998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,1,power_law_1.01,0.09494400024414062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,1,power_law_1.01,0.43029122352600097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,1,power_law_1.01,0.10038399696350098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,1,power_law_1.01,0.10690560340881347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,1,power_law_1.01,0.5292799949645997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,1,power_law_1.01,0.10359679460525513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,1,power_law_1.01,0.16904319524765016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,1,power_law_1.01,0.6201280117034912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,1,power_law_1.01,0.1672063946723938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,1,power_law_1.01,0.16718080043792724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,1,power_law_1.01,0.7768576145172119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,1,power_law_1.01,0.1696895956993103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,1,power_law_1.01,0.1689087986946106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,1,power_law_1.01,1.0156415939331054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,1,power_law_1.01,0.16550400257110595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,1,power_law_1.01,0.17136640548706056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,1,balanced,1.774197260538737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,1,power_law_1.01,1.2541440010070801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,1,power_law_1.01,0.1718783974647522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,1,power_law_1.01,0.16961920261383057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,1,power_law_1.01,1.7206207275390626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,1,power_law_1.01,0.16717439889907837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,1,power_law_1.01,0.16969599723815917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,1,power_law_1.01,2.230963134765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,1,power_law_1.01,0.1592512011528015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,1,power_law_1.01,0.17008639574050904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,1,power_law_1.01,0.16560640335083007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,1,power_law_1.01,3.1623615264892577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,1,power_law_1.01,0.16696959733963013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,1,power_law_1.01,0.16732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,1,power_law_1.01,0.1675328016281128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,1,power_law_1.01,4.1273857116699215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,1,power_law_1.01,0.1605568051338196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,1,power_law_1.01,0.1902143955230713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,1,power_law_1.01,5.123219299316406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,1,power_law_1.01,0.17580800056457518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,1,power_law_1.01,0.18300800323486327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,1,power_law_1.01,0.1840000033378601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,1,power_law_1.01,0.20616960525512695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,1,power_law_1.01,8.006489562988282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,1,power_law_1.01,0.1912832021713257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,1,power_law_1.01,0.20131840705871581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,1,power_law_1.01,0.20451200008392334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,1,power_law_1.01,0.23563520908355712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,1,power_law_1.01,16.11937255859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,1,power_law_1.01,0.21118719577789308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,1,power_law_1.01,0.26698238849639894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,1,balanced,3.5115038553873696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,1,power_law_1.01,0.25065600872039795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,1,power_law_1.01,0.31454079151153563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.01,0.27985920906066897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,1,power_law_1.01,0.3571903944015503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.01,0.34983038902282715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,1,power_law_1.01,0.44051198959350585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.01,0.40230398178100585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,1,power_law_1.01,0.587385606765747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.01,0.5364736080169678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,1,power_law_1.01,0.7152063846588135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.01,0.6543231964111328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,1,power_law_1.01,1.020742416381836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.01,0.8960000038146972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,1,power_law_1.01,1.3371392250061036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.01,1.1199872016906738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,1,power_law_1.01,1.923583984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.01,1.6802560806274414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,1,power_law_1.01,2.4954368591308596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.01,2.1761472702026365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,1,power_law_1.01,3.0419071197509764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.01,2.6350591659545897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,1,power_law_1.01,4.833804702758789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,1,power_law_1.01,4.180908966064453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,1,power_law_1.01,9.542176055908204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,1,power_law_1.01,8.315654754638672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,1,power_law_1.01,0.021401600539684297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,1,power_law_1.2,0.10569599866867066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,1,power_law_1.01,0.02401279956102371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,1,power_law_1.2,0.1415488004684448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,1,power_law_1.01,0.024486400187015533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,1,power_law_1.2,0.1788159966468811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,1,power_law_1.01,0.02545279860496521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,1,power_law_1.2,0.28750081062316896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,1,power_law_1.01,0.030272001028060914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,1,power_law_1.2,0.3002559900283813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,1,power_law_1.01,0.040166398882865904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,1,power_law_1.2,0.289305591583252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,1,power_law_1.01,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,1,power_law_1.01,0.038841599225997926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,1,power_law_1.2,0.29577600955963135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,1,power_law_1.01,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,1,power_law_1.2,0.29828479290008547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,1,power_law_1.01,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,1,power_law_1.2,0.2916543960571289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,1,power_law_1.01,0.044095999002456664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,1,power_law_1.2,0.2847743988037109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,1,power_law_1.01,0.0385343998670578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,1,power_law_1.2,0.29563519954681394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,1,power_law_1.01,0.04164479970932007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,1,power_law_1.2,0.290720009803772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,1,power_law_1.01,0.047366398572921756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,1,power_law_1.2,0.3076159954071045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,1,power_law_1.2,0.07108479738235474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,1,power_law_1.01,0.09994879961013795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,1,power_law_1.2,0.31965439319610595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,1,power_law_1.2,0.08324480056762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,1,power_law_1.2,0.09166079759597778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,1,power_law_1.01,0.10201599597930908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,1,power_law_1.2,0.35280001163482666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,1,power_law_1.2,0.08904320001602173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,1,power_law_1.01,0.12874239683151245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,1,power_law_1.2,0.10752639770507813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,1,power_law_1.2,0.3362047910690308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,1,power_law_1.2,0.10729600191116333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,1,power_law_1.01,0.0735871970653534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,1,power_law_1.2,0.1605631947517395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,1,power_law_1.2,0.3682559967041016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,1,power_law_1.2,0.16320639848709106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.01,0.09107840061187744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,1,power_law_1.2,0.16776959896087645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,1,power_law_1.2,0.45216641426086424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,1,power_law_1.2,0.16942720413208007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.01,0.1220736026763916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,1,power_law_1.2,0.17091200351715088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,1,power_law_1.2,0.49064321517944337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,1,power_law_1.2,0.16929279565811156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.01,0.14750720262527467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,1,power_law_1.2,0.6291007995605469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,1,power_law_1.2,0.16830079555511473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.01,0.2052095890045166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,1,power_law_1.2,0.7480000019073486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,1,power_law_1.2,0.17017600536346436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.01,0.25973761081695557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,1,power_law_1.2,0.9903552055358886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,1,power_law_1.2,0.16473599672317504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.01,0.3683199882507324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,1,power_law_1.2,1.2505855560302734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,1,power_law_1.2,0.1676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.01,0.47199358940124514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,1,power_law_1.2,1.7380992889404296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,1,power_law_1.2,0.16961920261383057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.01,0.6958784103393555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,1,power_law_1.2,0.17004799842834473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,1,power_law_1.2,2.183737564086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.01,0.9140607833862304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,1,power_law_1.2,0.1715648055076599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,1,power_law_1.2,3.168448066711426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,1,power_law_1.2,0.17643519639968872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.01,1.1265088081359864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,1,power_law_1.2,4.078713607788086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,1,power_law_1.2,0.19479680061340332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,1,power_law_1.01,1.7794048309326171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,1,power_law_1.2,5.055014419555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,1,power_law_1.2,0.19482879638671874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,1,power_law_1.01,3.508287811279297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,1,power_law_1.2,0.21941120624542237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,1,power_law_1.2,7.92468490600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,1,power_law_1.2,0.24461441040039061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.2,0.28354558944702146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,1,power_law_1.2,16.11864318847656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.2,0.34974079132080077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.2,0.39466879367828367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.2,0.5275968074798584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.2,0.6528575897216797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.2,0.8847552299499511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.2,1.153996753692627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.2,1.6309696197509767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.2,2.158060836791992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.2,2.6414207458496093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,1,power_law_1.2,4.155795288085938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,1,power_law_1.2,8.306982421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,1,power_law_1.2,0.021267199516296388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,1,power_law_1.2,0.024326400458812715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,1,power_law_1.2,0.0240447998046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,1,power_law_1.2,0.024928000569343568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,1,power_law_1.2,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,1,power_law_1.2,0.0401856005191803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,1,power_law_1.2,0.03457919955253601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,1,power_law_1.2,0.03968639969825745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,1,power_law_1.2,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,1,power_law_1.2,0.03660799860954285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,1,power_law_1.2,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,1,power_law_1.2,0.03944959938526153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,1,power_law_1.2,0.04078719913959503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,1,power_law_1.2,0.04649600088596344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,1,power_law_1.2,0.09992319941520691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,1,power_law_1.2,0.10187519788742065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,1,power_law_1.2,0.12861440181732178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,1,power_law_1.2,0.07357439994812012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.2,0.08931840062141419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.2,0.12114559412002564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.2,0.14529919624328613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.2,0.20389120578765868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.2,0.2572223901748657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.2,0.3671999931335449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.2,0.4747903823852539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.2,0.6890048027038574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.2,0.9127488136291504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.2,1.1285311698913574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,1,power_law_1.2,1.772275161743164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,1,power_law_1.2,3.5139583587646483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,4,balanced,0.09482666850090027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,4,balanced,0.12366400162378947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,4,balanced,0.2050079902013143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,4,balanced,0.33006399869918823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,4,balanced,0.5458933512369791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,4,balanced,1.0048906803131104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,4,balanced,1.0096533298492432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,4,balanced,1.0072906812032063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,4,balanced,1.0327893098195393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,4,balanced,1.0542933146158855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,1,power_law_1.2,0.16784000396728516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,4,balanced,1.019536018371582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,1,power_law_1.2,0.17150720357894897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,4,balanced,1.027359962463379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,1,power_law_1.2,0.16624000072479247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,4,balanced,1.0761706829071045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,1,power_law_1.2,0.1713152050971985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,4,balanced,0.985525369644165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,1,power_law_1.2,0.17375359535217286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,4,balanced,1.1323359807332356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,1,power_law_1.2,0.17231359481811523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,4,balanced,0.07089599967002869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,4,balanced,1.0218186378479004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,1,power_law_1.2,0.18529280424118041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,4,balanced,0.09011200070381165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,1,power_law_1.2,0.17984639406204223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,4,balanced,0.98744002978007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,4,balanced,0.12807466586430868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,1,power_law_1.2,0.2083967924118042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,4,balanced,0.2010186711947123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,4,balanced,1.0139520168304443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,1,power_law_1.2,0.19559680223464965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,4,balanced,0.3460693359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,1,power_law_1.2,0.22608640193939208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,4,balanced,0.5678293307622274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,4,balanced,1.017957369486491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,1,power_law_1.2,0.2713920116424561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,4,balanced,0.5672800143559774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,1,power_law_1.2,0.30595200061798095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,4,balanced,0.5458879868189493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,4,balanced,1.0175573031107585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,1,power_law_1.2,0.3622976064682007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,4,balanced,0.5473119815190634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,1,power_law_1.2,0.423635196685791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,4,balanced,0.5512799819310507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,1,power_law_1.2,0.5963327884674072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,4,balanced,0.5436746676762899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,4,balanced,1.0179839928944905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,1,power_law_1.2,0.733676815032959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,4,balanced,0.5497226715087891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,1,power_law_1.2,1.0168895721435547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,4,balanced,0.6073919932047526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,1,power_law_1.2,1.3214912414550781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,4,balanced,0.5614986817042033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,4,balanced,1.0515573024749756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,1,power_law_1.2,1.897702407836914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,4,balanced,0.5704960028330485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,1,power_law_1.2,2.495372772216797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,4,balanced,0.6127413511276245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,1,power_law_1.2,3.067385673522949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,4,balanced,0.5531520048777262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,1,power_law_1.2,4.799212646484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,4,balanced,1.1000906626383464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,4,balanced,0.5962293148040771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,1,power_law_1.2,9.488377380371094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,4,balanced,0.6061813433965048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,4,balanced,0.6100960175196329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,4,balanced,1.7616532643636067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,4,balanced,0.6267840067545573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,4,balanced,0.6798720359802246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,4,balanced,0.745568037033081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,4,balanced,1.6262399355570476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,4,balanced,1.1600533326466878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,4,balanced,2.2652319272359214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,4,balanced,1.1665173371632893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,4,balanced,1.629685401916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,4,balanced,2.9145867029825845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,4,balanced,2.0845813751220703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,4,balanced,3.5787680943806968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,4,balanced,2.5803732872009277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,4,balanced,0.07726400097211202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,4,balanced,0.10302399595578511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,4,balanced,0.14287466804186502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,4,balanced,0.23064533869425455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,4,balanced,0.3633280197779338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,4,balanced,0.5933333237965902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,4,balanced,0.5616586605707804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,4,balanced,0.5404213269551595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,4,balanced,0.5406186580657959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,4,balanced,0.5377600193023682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,4,balanced,0.5229066610336304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,4,balanced,0.4980693260828654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,4,balanced,0.5234293142954508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,4,balanced,0.554090658823649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,4,balanced,0.5433653195699056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,4,balanced,0.5706080198287964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,4,balanced,0.5378133455912272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,4,balanced,0.5947626829147339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,4,balanced,5.5933278401692705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,4,balanced,0.5692426760991415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,4,balanced,4.035749435424805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,4,balanced,0.5609279870986938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,4,balanced,0.5597173372904459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,4,balanced,0.585530678431193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,4,balanced,0.6276959975560507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,4,balanced,1.0887573560078938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,4,balanced,0.044490665197372437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,4,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,4,balanced,0.06391466657320659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,4,balanced,0.09248532851537068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,4,balanced,0.14125333229700723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,4,balanced,0.24548266331354776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,4,balanced,0.2571306626001994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,4,balanced,1.0505173206329346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,4,balanced,0.26849599679311115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,4,balanced,0.25563732782999676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,4,balanced,0.26232000192006427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,4,balanced,0.27508799235026044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,4,balanced,0.3009973367055257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,4,balanced,0.30075732866923016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,4,balanced,0.3068160017331441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,4,balanced,0.4249333143234253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,4,balanced,0.4224960009256999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,4,balanced,1.4646080334981282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,4,balanced,0.4251360098520915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,4,balanced,0.6365600029627482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,4,balanced,0.7100319862365723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,4,balanced,0.5965173244476318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,4,balanced,0.6382559935251871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,4,balanced,1.9038079579671223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,4,balanced,0.6951999664306641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,4,balanced,8.069674809773764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,4,balanced,11.14242172241211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,4,balanced,0.7615253130594889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,4,balanced,0.8731040159861246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,4,balanced,2.323967933654785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,4,balanced,0.989786704381307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,4,balanced,1.6607252756754558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,4,balanced,1.8570559819539387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,4,balanced,3.630805333455404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,4,balanced,2.5393706957499185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,4,balanced,3.587909380594889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,4,balanced,7.1600799560546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,4,power_law_1.01,0.20229120254516603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,4,power_law_1.01,0.3424639940261841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,4,power_law_1.01,0.21440000534057618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,4,power_law_1.01,0.13349119424819947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,4,power_law_1.01,0.28170878887176515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,4,power_law_1.01,0.20942718982696534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,4,power_law_1.01,0.4231679916381836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,4,power_law_1.01,0.15215359926223754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,4,power_law_1.01,0.562278413772583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,4,power_law_1.01,0.17976319789886475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,4,power_law_1.01,0.7509119987487793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,4,power_law_1.01,0.27130239009857177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,4,power_law_1.01,0.813593578338623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,4,power_law_1.01,0.3236991882324219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,4,power_law_1.01,0.8094719886779785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,4,power_law_1.01,0.40885119438171386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,4,power_law_1.01,0.8624768257141113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,4,power_law_1.01,0.45072641372680666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,4,balanced,7.100528081258138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,4,power_law_1.01,0.8801152229309082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,4,power_law_1.01,0.45833601951599123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,4,power_law_1.01,0.9017215728759765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,4,power_law_1.01,0.4654079914093018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,4,power_law_1.01,0.8930879592895508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,4,power_law_1.01,0.4854015827178955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,4,power_law_1.01,0.8924799919128418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,4,power_law_1.01,0.5011903762817382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,4,power_law_1.01,0.9477439880371094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,4,power_law_1.01,0.5091648101806641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,4,power_law_1.01,0.8942591667175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,4,power_law_1.01,0.5158592224121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,4,power_law_1.01,0.9112256050109864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,4,power_law_1.01,0.5207039833068847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,4,power_law_1.01,0.9757375717163086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,4,power_law_1.01,0.5305024147033691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,4,power_law_1.01,1.031436824798584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,4,power_law_1.01,0.5368000030517578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,4,power_law_1.01,1.1471487998962402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,4,power_law_1.01,0.5921599864959717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,4,power_law_1.01,1.2805695533752441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,4,power_law_1.01,0.6283135890960694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,4,power_law_1.01,1.5498432159423827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,4,power_law_1.01,0.6813119888305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,4,power_law_1.01,1.700396728515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,4,power_law_1.01,0.7668928146362305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,4,power_law_1.01,2.1991039276123048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,4,power_law_1.01,0.9219200134277343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,4,power_law_1.01,2.501683235168457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,4,power_law_1.01,1.1123071670532227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,4,power_law_1.01,3.569139099121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,4,power_law_1.01,1.3961919784545898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,4,power_law_1.01,4.439718246459961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,4,power_law_1.01,1.6353216171264648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,4,power_law_1.01,5.213452911376953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,4,power_law_1.01,2.2633855819702147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,4,power_law_1.01,7.964575958251953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,4,power_law_1.01,2.8933055877685545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,4,power_law_1.01,3.4456192016601563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,4,power_law_1.01,14.56534423828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,4,power_law_1.01,5.4821311950683596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,4,power_law_1.01,9.794080352783203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,4,power_law_1.01,0.1519551992416382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,4,power_law_1.01,0.22368640899658204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,4,power_law_1.01,0.12433279752731323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,4,power_law_1.01,0.1821887969970703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,4,power_law_1.01,0.2663232088088989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,4,power_law_1.01,0.33438079357147216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,4,power_law_1.01,0.4180736064910889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,4,power_law_1.01,0.4488831996917725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,4,power_law_1.01,0.46944642066955566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,4,power_law_1.01,0.4779776096343994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,4,power_law_1.01,0.4887807846069336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,4,power_law_1.01,0.4885824203491211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,4,power_law_1.01,0.49152002334594724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,4,power_law_1.01,0.4947648048400879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,4,power_law_1.01,0.5155327796936036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,4,power_law_1.01,0.517907190322876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,4,power_law_1.01,0.530617618560791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,4,power_law_1.01,0.5595136165618897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.01,0.6029823780059814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.01,0.6652607917785645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.01,0.719488000869751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.01,0.8932607650756836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.01,0.9962688446044922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.01,1.3574015617370605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.01,1.5925567626953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.01,2.07256965637207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.01,2.605356788635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,4,power_law_1.01,0.04826880097389221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,4,power_law_1.01,3.1699712753295897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,4,power_law_1.01,0.07517439723014832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,4,power_law_1.01,0.06037120223045349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,4,power_law_1.01,4.921964645385742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,4,power_law_1.01,0.08567039966583252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,4,power_law_1.01,0.10448640584945679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,4,power_law_1.01,9.890898895263671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,4,power_law_1.01,0.14629119634628296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,4,power_law_1.01,0.18328959941864015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,4,power_law_1.01,0.1984063982963562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,4,power_law_1.01,0.20004479885101317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,4,power_law_1.01,0.20276479721069335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,4,power_law_1.01,0.2120896100997925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,4,power_law_1.01,0.2156287908554077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,4,power_law_1.01,0.21320960521697999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,4,power_law_1.01,0.22371840476989746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,4,power_law_1.01,0.26935040950775146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,4,power_law_1.01,0.25160961151123046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,4,power_law_1.01,0.2666879892349243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,4,power_law_1.01,0.5204864025115967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.01,0.5592000007629394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.01,0.42040319442749025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,4,power_law_1.2,0.13038079738616942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.01,0.45408000946044924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,4,power_law_1.2,0.20023040771484374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.01,0.4835008144378662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,4,power_law_1.2,0.11706880331039429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,4,power_law_1.2,0.1920575976371765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.01,0.5758975982666016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,4,power_law_1.2,0.1649791955947876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,4,power_law_1.2,0.3183808088302612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.01,0.6891520023345947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,4,power_law_1.2,0.23438079357147218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,4,power_law_1.2,0.17487360239028932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.01,0.8333312034606933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,4,power_law_1.2,0.27118721008300783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,4,power_law_1.2,0.24658560752868652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.01,1.1070976257324219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,4,power_law_1.2,0.41095681190490724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,4,power_law_1.2,0.42110719680786135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.01,1.4235456466674805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,4,power_law_1.2,0.4341760158538818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,4,power_law_1.2,0.5183487892150879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,4,power_law_1.01,1.8305023193359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,4,power_law_1.2,0.46280322074890134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,4,power_law_1.2,0.73537278175354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,4,power_law_1.01,2.808723258972168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,4,power_law_1.2,0.4757887840270996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,4,power_law_1.2,0.7891007900238037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,4,power_law_1.2,0.4698431968688965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,4,power_law_1.2,0.8031295776367188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,4,power_law_1.01,5.1094718933105465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,4,power_law_1.2,0.485152006149292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,4,power_law_1.2,0.809331226348877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,4,power_law_1.2,0.5001088142395019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,4,power_law_1.2,0.8827008247375489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,4,power_law_1.2,0.517574405670166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,4,power_law_1.2,0.9175871849060059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,4,power_law_1.2,0.5348095893859863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,4,power_law_1.2,0.8770432472229004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,4,power_law_1.2,0.5449600219726562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,4,power_law_1.2,0.9095871925354004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,4,power_law_1.2,0.5661312103271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,4,power_law_1.2,0.9687935829162597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,4,power_law_1.2,0.6173439979553222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,4,power_law_1.2,0.9263296127319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,4,power_law_1.2,0.6542272090911865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,4,power_law_1.2,0.9724736213684082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,4,power_law_1.2,0.7559103965759277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,4,power_law_1.2,1.0524607658386231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,4,power_law_1.2,0.8181952476501465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,4,power_law_1.2,1.130777645111084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,4,power_law_1.2,1.0049280166625976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,4,power_law_1.2,1.2843135833740233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,4,power_law_1.2,1.1575488090515136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,4,power_law_1.2,1.4166784286499023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,4,power_law_1.2,1.403219223022461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,4,power_law_1.2,1.6184576034545899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,4,power_law_1.2,1.6918912887573243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,4,power_law_1.2,1.9223424911499023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,4,power_law_1.2,2.352742385864258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,4,power_law_1.2,2.3511104583740234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,4,power_law_1.2,3.000332832336426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,4,power_law_1.2,2.8235008239746096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,4,power_law_1.2,3.7880382537841797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,4,power_law_1.2,3.8379135131835938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,4,power_law_1.2,5.705145645141601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,4,power_law_1.2,5.260550308227539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,4,power_law_1.2,10.212236785888672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,4,power_law_1.2,6.052147293090821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,4,power_law_1.2,9.022112274169922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,4,power_law_1.2,16.57771453857422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,4,power_law_1.2,0.1471743941307068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,4,power_law_1.2,0.20608000755310057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,4,power_law_1.2,0.11781120300292969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,4,power_law_1.2,0.1704576015472412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,4,power_law_1.2,0.2483135938644409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,4,power_law_1.2,0.32113919258117674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,4,power_law_1.2,0.44671359062194826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,4,power_law_1.2,0.4677440166473389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,4,power_law_1.2,0.4803775787353516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,4,power_law_1.2,0.47097601890563967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,4,power_law_1.2,0.4782271862030029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,4,power_law_1.2,0.4823296070098877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,4,power_law_1.2,0.5112832069396973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,4,power_law_1.2,0.5132160186767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,4,power_law_1.2,0.5088831901550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,4,power_law_1.2,0.5406400203704834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,4,power_law_1.2,0.5456384181976318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,4,power_law_1.2,0.5965375900268555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.2,0.6417727947235108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.2,0.7121280193328857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.2,0.8038784027099609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.2,0.9308671951293945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.2,1.115187168121338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.2,1.345024013519287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.2,1.607686424255371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.2,2.12806396484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,4,power_law_1.2,0.04774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.2,2.690278434753418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,4,power_law_1.2,0.07475200295448303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,4,power_law_1.2,3.469798278808594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,4,power_law_1.2,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,4,power_law_1.2,5.394060897827148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,4,power_law_1.2,0.07624319791793824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,4,power_law_1.2,0.0945855975151062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,4,power_law_1.2,9.916422271728516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,4,power_law_1.2,0.12102400064468384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,4,power_law_1.2,0.17137279510498046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,4,power_law_1.2,0.18951679468154908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,4,power_law_1.2,0.20278398990631102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,4,power_law_1.2,0.20964479446411133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,4,power_law_1.2,0.20165760517120362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,4,power_law_1.2,0.22145280838012696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,4,power_law_1.2,0.22658560276031495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,4,power_law_1.2,0.2298367977142334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,4,power_law_1.2,0.2750080108642578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,4,power_law_1.2,0.25813119411468505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,4,power_law_1.2,0.27800960540771485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,4,power_law_1.2,0.5228415966033936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.2,0.592851209640503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.2,0.45542402267456056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,2,balanced,0.09950400392214458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.2,0.4578239917755127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,2,balanced,0.12447999914487202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.2,0.5537471771240234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,2,balanced,0.18837867180506387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,2,balanced,0.3250346581141154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.2,0.5902527809143067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,2,balanced,0.543018658955892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.2,0.7242688179016114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,2,balanced,0.9730186462402344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,2,balanced,0.9565172990163168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.2,0.8709759712219238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,2,balanced,0.9640533129374186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.2,1.2561856269836427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,2,balanced,1.0010453065236409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,2,balanced,0.9969440301259359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.2,1.5338944435119628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,2,balanced,0.07278400162855785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,2,balanced,1.0055306752522786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,4,power_law_1.2,1.9083711624145507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,2,balanced,0.09011200070381165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,2,balanced,0.98362135887146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,2,balanced,0.127210666735967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,2,balanced,1.0186346371968586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,4,power_law_1.2,2.7781824111938476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,2,balanced,0.20898133516311646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,2,balanced,1.0097333590189617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,2,balanced,0.3209120035171509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,4,power_law_1.2,5.456659317016602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,2,balanced,0.5580960114796957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,2,balanced,0.9945066769917806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,2,balanced,0.5446826616923014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,2,balanced,1.0345653692881267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,2,balanced,0.5250293413798014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,2,balanced,0.5386079947153727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,2,balanced,0.9747200012207031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,2,balanced,0.5371946493784586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,2,balanced,1.0507573286692302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,2,balanced,0.5392586787541708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,2,balanced,0.5282346804936727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,2,balanced,1.0183093547821045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,2,balanced,0.556389331817627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,2,balanced,0.5525759855906168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,2,balanced,1.04147203763326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,2,balanced,0.5589173237482706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,2,balanced,0.5886666774749756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,2,balanced,1.0577279726664226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,2,balanced,0.5804479916890463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,2,balanced,0.6028800010681152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,2,balanced,1.1330986817677815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,2,balanced,0.606661319732666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,2,balanced,0.6492266654968262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,2,balanced,1.2162933349609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,2,balanced,0.6798559824625651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,2,balanced,0.7839306990305582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,2,balanced,1.8535839716593425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,2,balanced,0.8825013637542725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,2,balanced,1.8245760599772136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,2,balanced,1.3096960385640461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,2,balanced,1.4196906089782715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,2,balanced,2.549071947733561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,2,balanced,1.9935466448465984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,2,balanced,3.235472043355306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,2,balanced,2.5717493693033853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,2,balanced,0.08728532989819844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,2,balanced,0.11272000273068745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,2,balanced,3.9474080403645835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,2,balanced,0.1543786625067393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,2,balanced,0.22530666987101236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,2,balanced,0.36749335130055744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,2,balanced,0.601418654123942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,2,balanced,0.5830026865005493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,2,balanced,0.5655680100123087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,2,balanced,3.160991986592611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,2,balanced,0.5528266827265421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,2,balanced,0.5490986506144205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,2,balanced,0.5360000133514404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,2,balanced,0.5285919904708862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,2,balanced,0.5380320151646932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,2,balanced,0.5328799883524576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,2,balanced,0.5403039852778116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,2,balanced,0.5783573389053345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,2,balanced,0.5632693370183309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,2,balanced,0.5958026647567749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,2,balanced,0.5833280086517334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,2,balanced,0.6154346863428751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,2,balanced,0.029850666721661884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,2,balanced,0.6255946556727091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,2,balanced,0.04474133253097534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,2,balanced,6.387957255045573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,2,balanced,0.05884266893068949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,2,balanced,0.08402132987976074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,2,balanced,0.14033066232999167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,2,balanced,0.7039733727773031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,2,balanced,0.24219733476638794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,2,balanced,5.063808123270671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,2,balanced,0.24734934171040854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,2,balanced,0.25376532475153607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,2,balanced,0.25942399104436237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,2,balanced,0.7943680286407471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,2,balanced,0.2683200041453044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,2,balanced,0.2820746699968974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,2,balanced,0.26758400599161786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,2,balanced,0.27426666021347046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,2,balanced,0.2865973313649495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,2,balanced,1.2670986652374268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,2,balanced,0.29688533147176105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,2,balanced,0.2988373239835103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,2,balanced,0.3054933349291484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,2,balanced,0.3963306744893392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,2,balanced,0.3866293430328369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,2,balanced,1.3514986038208008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,2,balanced,0.6991413434346517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,2,balanced,0.7570292949676514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,2,balanced,0.6349066495895386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,2,balanced,1.902901331583659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,2,balanced,0.673087994257609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,2,balanced,0.7243253389994303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,2,balanced,2.497722625732422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,2,balanced,0.7960373560587565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,2,balanced,12.957220713297525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,2,balanced,10.096373240152994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,2,balanced,1.3447252909342449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,2,balanced,3.0642080307006836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,2,balanced,1.4718613624572754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,2,balanced,2.0290986696879068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,2,balanced,4.815247853597005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,2,balanced,2.8152265548706055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,2,balanced,9.528805414835611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,2,balanced,5.569850921630859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,2,power_law_1.01,0.12359039783477783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,2,power_law_1.01,0.1983423948287964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,2,power_law_1.01,0.1714367985725403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,2,power_law_1.01,0.24357120990753173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,2,power_law_1.01,0.3887871980667114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,2,power_law_1.01,0.5133312225341797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,2,power_law_1.01,0.09136639833450318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,2,power_law_1.01,0.6999423980712891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,2,power_law_1.01,0.1314239978790283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,2,power_law_1.01,0.7460671901702881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,2,power_law_1.01,0.11170560121536255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,2,power_law_1.01,0.15975680351257324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,2,power_law_1.01,0.7740416049957275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,2,power_law_1.01,0.24533119201660156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,2,power_law_1.01,0.7699007987976074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,2,power_law_1.01,0.297273588180542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,2,power_law_1.01,0.8243840217590332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,2,power_law_1.01,0.4087423801422119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,2,power_law_1.01,0.8286975860595703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,2,power_law_1.01,0.4206528186798096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,2,power_law_1.01,0.8358016014099121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,2,power_law_1.01,0.4412799835205078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,2,power_law_1.01,0.8726207733154296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,2,power_law_1.01,0.4441215991973877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,2,power_law_1.01,0.8990719795227051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,2,power_law_1.01,0.4682432174682617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,2,power_law_1.01,0.9043968200683594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,2,power_law_1.01,0.47358717918396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,2,power_law_1.01,0.9191424369812011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,2,power_law_1.01,0.47922558784484864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,2,power_law_1.01,0.9776703834533691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,2,power_law_1.01,0.5030335903167724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,2,power_law_1.01,1.005568027496338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,2,power_law_1.01,0.5139904022216797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,2,power_law_1.01,1.154207992553711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,2,power_law_1.01,0.519161605834961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,2,power_law_1.01,1.2611200332641601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,2,power_law_1.01,0.531718397140503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,2,power_law_1.01,1.4554495811462402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,2,power_law_1.01,0.5786623954772949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,2,power_law_1.01,1.6979904174804688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,2,power_law_1.01,0.6269504070281983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,2,power_law_1.01,2.078713607788086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,2,power_law_1.01,0.7034815788269043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,2,power_law_1.01,2.6325439453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,2,power_law_1.01,0.7865280151367188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,2,power_law_1.01,3.526457595825195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,2,power_law_1.01,0.929747200012207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,2,power_law_1.01,4.211507034301758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,2,power_law_1.01,1.1053248405456544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,2,power_law_1.01,5.1284736633300785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,2,power_law_1.01,1.4525823593139648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,2,power_law_1.01,1.7959871292114258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,2,power_law_1.01,7.820735931396484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,2,power_law_1.01,2.475200080871582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,2,power_law_1.01,15.178585815429688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,2,power_law_1.01,3.0690431594848633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,2,power_law_1.01,3.7702400207519533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,2,power_law_1.01,5.608031845092773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,2,power_law_1.01,10.815897369384766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,2,power_law_1.01,0.11364480257034301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,2,power_law_1.01,0.14629119634628296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,2,power_law_1.01,0.12188800573348998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,2,power_law_1.01,0.17946239709854125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,2,power_law_1.01,0.2600640058517456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,2,power_law_1.01,0.3446719884872437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,2,power_law_1.01,0.4160639762878418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,2,power_law_1.01,0.43912320137023925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,2,power_law_1.01,0.02813439965248108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,2,power_law_1.01,0.4468480110168457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,2,power_law_1.01,0.048390400409698484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,2,power_law_1.01,0.0489984005689621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,2,power_law_1.01,0.4685311794281006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,2,power_law_1.01,0.06997759938240052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,2,power_law_1.01,0.47246079444885253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,2,power_law_1.01,0.0964735984802246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,2,power_law_1.01,0.48145279884338377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,2,power_law_1.01,0.1297152042388916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,2,power_law_1.01,0.49869441986083984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,2,power_law_1.01,0.17587200403213502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,2,power_law_1.01,0.5250879764556885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,2,power_law_1.01,0.18389760255813598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,2,power_law_1.01,0.5159167766571044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,2,power_law_1.01,0.19347840547561646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,2,power_law_1.01,0.5261951923370362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,2,power_law_1.01,0.19247360229492189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,2,power_law_1.01,0.5411200046539306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,2,power_law_1.01,0.20944640636444092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,2,power_law_1.01,0.5765056133270263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,2,power_law_1.01,0.2076416015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.01,0.6188032150268554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,2,power_law_1.01,0.21410560607910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.01,0.6784639835357666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,2,power_law_1.01,0.2246335983276367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.01,0.7634111881256104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,2,power_law_1.01,0.22855041027069092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.01,0.9383808135986328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,2,power_law_1.01,0.2360703945159912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.01,1.0836095809936523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,2,power_law_1.01,0.24649600982666015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.01,1.3852479934692383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,2,power_law_1.01,0.2542464017868042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.01,1.6970624923706055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.01,0.27952640056610106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.01,2.4041343688964845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.01,0.5853568077087402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.01,2.868716812133789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.01,0.6644991874694824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.01,0.4830207824707031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,2,power_law_1.01,3.7258495330810546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.01,0.49906558990478517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,2,power_law_1.01,5.689459228515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.01,0.6335040092468261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.01,0.7615359783172607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,2,power_law_1.01,10.858598327636718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.01,0.9862079620361328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.01,1.2711551666259766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,2,power_law_1.01,1.5265024185180665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,2,power_law_1.01,2.366502380371094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,2,power_law_1.01,4.3167167663574215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,2,power_law_1.2,0.1259392023086548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,2,power_law_1.2,0.17758079767227172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,2,power_law_1.2,0.1425536036491394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,2,power_law_1.2,0.2554239988327026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,2,power_law_1.2,0.3464384078979492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,2,power_law_1.2,0.46694397926330566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,2,power_law_1.2,0.6934207916259766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,2,power_law_1.2,0.7302720069885253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,2,power_law_1.2,0.7675392150878906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,2,power_law_1.2,0.7589119911193848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,2,power_law_1.2,0.8340543746948242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,2,power_law_1.2,0.8582207679748535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,2,power_law_1.2,0.09108480215072631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,2,power_law_1.2,0.8766079902648926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,2,power_law_1.2,0.13022719621658324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,2,power_law_1.2,0.11450879573822022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,2,power_law_1.2,0.8701184272766114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,2,power_law_1.2,0.17041280269622802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,2,power_law_1.2,0.8969216346740723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,2,power_law_1.2,0.2284032106399536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,2,power_law_1.2,0.9507519721984863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,2,power_law_1.2,0.2774463891983032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,2,power_law_1.2,0.9675840377807617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,2,power_law_1.2,0.40546560287475586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,2,power_law_1.2,1.0218432426452637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,2,power_law_1.2,0.419539213180542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,2,power_law_1.2,1.0680319786071777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,2,power_law_1.2,0.43486719131469725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,2,power_law_1.2,1.2027263641357422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,2,power_law_1.2,0.44173440933227537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,2,power_law_1.2,1.3432640075683593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,2,power_law_1.2,0.47545599937438965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,2,power_law_1.2,1.5530303955078124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,2,power_law_1.2,0.47466239929199217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,2,power_law_1.2,1.8075199127197266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,2,power_law_1.2,0.500704002380371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,2,power_law_1.2,2.1381120681762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,2,power_law_1.2,0.5095424175262451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,2,power_law_1.2,2.681292724609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,2,power_law_1.2,0.5363455772399902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,2,power_law_1.2,3.760166549682617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,2,power_law_1.2,0.5433728218078613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,2,power_law_1.2,0.5684480190277099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,2,power_law_1.2,4.466259384155274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,2,power_law_1.2,0.6028927803039551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,2,power_law_1.2,5.2127422332763675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,2,power_law_1.2,0.650707197189331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,2,power_law_1.2,0.7730495929718018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,2,power_law_1.2,8.268230438232422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,2,power_law_1.2,0.8122879981994628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,2,power_law_1.2,0.9738176345825196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,2,power_law_1.2,14.732633972167969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,2,power_law_1.2,1.202348804473877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,2,power_law_1.2,1.4694272041320802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,2,power_law_1.2,1.8440191268920898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,2,power_law_1.2,2.571244812011719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,2,power_law_1.2,3.1894399642944338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,2,power_law_1.2,3.6245376586914064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,2,power_law_1.2,5.860383987426758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,2,power_law_1.2,11.33694076538086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,2,power_law_1.2,0.11225600242614746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,2,power_law_1.2,0.14324480295181274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,2,power_law_1.2,0.10966399908065796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,2,power_law_1.2,0.028255999088287354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,2,power_law_1.2,0.1661631941795349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,2,power_law_1.2,0.048147198557853696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,2,power_law_1.2,0.24487679004669188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,2,power_law_1.2,0.043968001008033754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,2,power_law_1.2,0.29529600143432616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,2,power_law_1.2,0.06682239770889283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,2,power_law_1.2,0.43669757843017576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,2,power_law_1.2,0.0961023986339569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,2,power_law_1.2,0.45903358459472654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,2,power_law_1.2,0.12083200216293336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,2,power_law_1.2,0.45206398963928224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,2,power_law_1.2,0.16631040573120118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,2,power_law_1.2,0.46871042251586914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,2,power_law_1.2,0.18059519529342652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,2,power_law_1.2,0.4762112140655518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,2,power_law_1.2,0.5024256229400634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,2,power_law_1.2,0.18802560567855836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,2,power_law_1.2,0.1932736039161682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,2,power_law_1.2,0.5163328170776367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,2,power_law_1.2,0.20979199409484864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,2,power_law_1.2,0.5314367771148681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,2,power_law_1.2,0.5318272113800049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,2,power_law_1.2,0.20765440464019774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,2,power_law_1.2,0.2194432020187378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,2,power_law_1.2,0.5351615905761719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,2,power_law_1.2,0.23123838901519775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,2,power_law_1.2,0.5602943897247314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,2,power_law_1.2,0.23374719619750978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,2,power_law_1.2,0.6097280025482178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,2,power_law_1.2,0.2342144012451172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.2,0.6504576206207275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,2,power_law_1.2,0.2571455955505371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.2,0.7233151912689209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,2,power_law_1.2,0.26966400146484376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.2,0.818841552734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.2,0.29234559535980226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.2,0.9868991851806641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.2,0.6151999950408935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.2,1.1265664100646973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.2,0.7063168048858642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.2,1.4482239723205566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.2,0.4957695960998535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.2,1.7596223831176758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.2,0.5515776157379151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.2,2.6593151092529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.2,0.6617792129516602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.2,3.0831296920776365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.2,0.8081279754638672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,2,power_law_1.2,3.6101310729980467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.2,1.0988351821899414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,2,power_law_1.2,5.593920135498047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.2,1.3762432098388673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,2,power_law_1.2,1.533407974243164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,2,power_law_1.2,11.091455841064453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,2,power_law_1.2,2.4938495635986326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,2,power_law_1.2,4.653542327880859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,1,balanced,0.1033066709836324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,1,balanced,0.12956800063451132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,1,balanced,0.1949066718419393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,1,balanced,0.33132266998291016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,1,balanced,0.5334186553955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,1,balanced,0.9732000033060709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,1,balanced,0.9781920115152994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,1,balanced,0.9726560115814209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,1,balanced,0.9831360181172689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,1,balanced,0.9896799723307291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,1,balanced,0.9604640007019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,1,balanced,0.9638933340708414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,1,balanced,0.9908746878306071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,1,balanced,0.08318933347860973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,1,balanced,0.9608853658040365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,1,balanced,0.10166399677594502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,1,balanced,0.9823199907938639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,1,balanced,0.13783466815948486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,1,balanced,0.21586666504542032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,1,balanced,0.9777119954427084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,1,balanced,0.3267040054003398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,1,balanced,0.5651413202285767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,1,balanced,0.9966026941935221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,1,balanced,0.562549352645874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,1,balanced,1.063968022664388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,1,balanced,0.5558506647745768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,1,balanced,0.5622773170471191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,1,balanced,0.5568053325017294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,1,balanced,1.0538346767425537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,1,balanced,0.5645386775334676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,1,balanced,0.5670933326085409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,1,balanced,1.124618689219157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,1,balanced,0.593450665473938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,1,balanced,0.5780160029729208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,1,balanced,1.1843360265096028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,1,balanced,0.5928053458531698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,1,balanced,0.6179893414179484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,1,balanced,0.6128319899241129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,1,balanced,1.326202630996704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,1,balanced,0.660368005434672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,1,balanced,0.6860799789428711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,1,balanced,0.7688266436258951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,1,balanced,1.4585919380187988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,1,balanced,0.8344533443450928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,1,balanced,0.989962657292684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,1,balanced,2.108282725016276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,1,balanced,1.1455787022908528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,1,balanced,1.6999252637227376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,1,balanced,2.3041653633117676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,1,balanced,1.9705066680908203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,1,balanced,3.261119842529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,1,balanced,2.8114986419677734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,1,balanced,4.129397392272949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,1,balanced,3.618069330851237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,1,balanced,0.02386133372783661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,1,balanced,0.029088000456492107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,1,balanced,0.04018666595220566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,1,balanced,0.05585599939028422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,1,balanced,0.08574400345484416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,1,balanced,0.14735999703407288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,1,balanced,0.14983466267585754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,1,balanced,0.15476800004641214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,1,balanced,0.10587732990582784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,1,balanced,0.1593653361002604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,1,balanced,0.1288586656252543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,1,balanced,0.1620746652285258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,1,balanced,0.17032533884048462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,1,balanced,4.449333190917969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,1,balanced,0.25732266902923584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,1,balanced,0.1705066760381063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,1,balanced,0.406607985496521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,1,balanced,0.17847466468811035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,1,balanced,0.6803680260976156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,1,balanced,0.18756800889968872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,1,balanced,5.033573468526204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,1,balanced,0.6588000059127808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,1,balanced,0.20603734254837036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,1,balanced,0.6530559857686361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,1,balanced,0.19234667221705118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,1,balanced,0.637664000193278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,1,balanced,0.1995733380317688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,1,balanced,0.627461314201355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,1,balanced,0.6290773153305054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,1,balanced,0.21802133321762085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,1,balanced,0.6285013357798258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,1,balanced,0.21875733137130737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,1,balanced,0.6287839810053507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,1,balanced,0.24064000447591147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,1,balanced,0.6551733414332072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,1,balanced,0.6528053283691406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,1,balanced,0.24926400184631348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,1,balanced,0.6809173425038656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,1,balanced,0.673957347869873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,1,balanced,0.2720800042152405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,1,balanced,0.7142240206400553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,1,balanced,0.7239840030670166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,1,balanced,0.5182506640752157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,1,balanced,0.7927626768747965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,1,balanced,7.004101435343425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,1,balanced,0.611135999361674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,1,balanced,0.8405386606852213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,1,balanced,0.5006346702575684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,1,balanced,0.9778986771901449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,1,balanced,7.8893280029296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,1,balanced,1.1216373443603516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,1,balanced,0.55894935131073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,1,balanced,1.7683040301005046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,1,balanced,0.9127039909362793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,1,balanced,2.011840025583903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,1,balanced,1.027743975321452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,1,balanced,2.852346738179525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,1,balanced,13.986778259277344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,1,balanced,1.3939839998881023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,1,balanced,3.7585652669270835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,1,balanced,16.248240152994793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,1,balanced,4.610650698343913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,1,balanced,1.9956960678100586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,1,balanced,7.32366943359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,1,balanced,3.91155211130778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,1,balanced,14.570154825846354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,1,power_law_1.01,0.09873920083045959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,1,power_law_1.01,0.12239359617233277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,1,power_law_1.01,0.15529600381851197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,1,power_law_1.01,0.261190390586853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,1,power_law_1.01,0.38601601123809814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,1,power_law_1.01,0.48581762313842775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,1,power_law_1.01,0.6849279880523682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,1,power_law_1.01,0.7199103832244873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,1,power_law_1.01,0.07697280049324036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,1,power_law_1.01,0.7421376228332519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,1,power_law_1.01,0.0961471974849701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,1,power_law_1.01,0.7468736171722412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,1,power_law_1.01,0.09716479778289795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,1,power_law_1.01,0.7977280139923095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,1,power_law_1.01,0.1776128053665161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,1,power_law_1.01,0.8257023811340332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,1,power_law_1.01,0.23563520908355712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,1,power_law_1.01,0.8472895622253418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,1,power_law_1.01,0.3010560035705566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,1,power_law_1.01,0.8621503829956054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,1,power_law_1.01,0.41032958030700684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,1,power_law_1.01,0.8971136093139649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,1,power_law_1.01,0.41948161125183103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,1,power_law_1.01,0.8911295890808105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,1,power_law_1.01,0.446124792098999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,1,power_law_1.01,0.9160191535949707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,1,power_law_1.01,0.4493311882019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,1,power_law_1.01,0.9704575538635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,1,power_law_1.01,0.4811583995819092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,1,power_law_1.01,1.0453760147094726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,1,power_law_1.01,0.474944019317627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,1,power_law_1.01,1.161952018737793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,1,power_law_1.01,0.4790783882141113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,1,power_law_1.01,1.295564842224121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,1,power_law_1.01,0.5152639865875244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,1,power_law_1.01,1.5273280143737793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,1,power_law_1.01,0.5294400215148926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,1,power_law_1.01,1.7666496276855468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,1,power_law_1.01,0.5353343963623047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,1,power_law_1.01,2.2733760833740235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,1,power_law_1.01,0.5633152008056641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,1,power_law_1.01,2.788697624206543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,1,power_law_1.01,0.6121280193328857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,1,power_law_1.01,3.7700286865234376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,1,power_law_1.01,0.6726528167724609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,1,power_law_1.01,0.7817471981048584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,1,power_law_1.01,4.788332748413086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,1,power_law_1.01,0.8884672164916992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,1,power_law_1.01,5.742784118652343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,1,power_law_1.01,1.0891839981079101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,1,power_law_1.01,8.657888031005859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,1,power_law_1.01,1.2975104331970215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,1,power_law_1.01,1.7163904190063477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,1,power_law_1.01,16.829656982421874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,1,power_law_1.01,2.126041603088379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,1,power_law_1.01,2.9605056762695314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,1,power_law_1.01,3.796908950805664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,1,power_law_1.01,4.643788909912109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,1,power_law_1.01,7.140563201904297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,1,power_law_1.01,13.933094787597657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,1,power_law_1.01,0.09971839785575867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,1,power_law_1.01,0.12926080226898193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,1,power_law_1.01,0.1362239956855774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,1,power_law_1.01,0.21223680973052977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,1,power_law_1.01,0.29178240299224856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,1,power_law_1.01,0.3542527914047241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,1,power_law_1.01,0.48423042297363283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,1,power_law_1.01,0.49529600143432617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,1,power_law_1.01,0.5047488212585449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,1,power_law_1.01,0.5225344181060791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,1,power_law_1.01,0.5309951782226563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,1,power_law_1.01,0.5481088161468506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,1,power_law_1.01,0.5680895805358886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,1,power_law_1.01,0.5668928146362304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,1,power_law_1.01,0.5719552040100098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,1,power_law_1.01,0.588915205001831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,1,power_law_1.01,0.6077119827270507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,1,power_law_1.01,0.6762495994567871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.01,0.7263360023498535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.01,0.8397120475769043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,1,power_law_1.01,0.022393600642681123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.01,0.9363391876220704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,1,power_law_1.01,0.028172799944877626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,1,power_law_1.01,0.03386879861354828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.01,1.1561280250549317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,1,power_law_1.01,0.04959999918937683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.01,1.3731583595275878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,1,power_law_1.01,0.08107519745826722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.01,1.8339199066162108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,1,power_law_1.01,0.12390400171279907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.01,2.2695615768432615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,1,power_law_1.01,0.15240960121154784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.01,3.142406463623047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,1,power_law_1.01,0.16749440431594848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.01,4.0050304412841795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,1,power_law_1.01,0.17377279996871947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.01,4.867916870117187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,1,power_law_1.01,0.1840127944946289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,1,power_law_1.01,0.1956544041633606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,1,power_law_1.01,7.466560363769531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,1,power_law_1.01,0.20342400074005126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,1,power_law_1.01,0.21395840644836425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,1,power_law_1.01,14.432102966308594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,1,power_law_1.01,0.23515520095825196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,1,power_law_1.01,0.22923519611358642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,1,power_law_1.01,0.23665919303894042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,1,power_law_1.01,0.25793280601501467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,1,power_law_1.01,0.26213760375976564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.01,0.28448638916015623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.01,0.30703999996185305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.01,0.33710079193115233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.01,0.7220032215118408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.01,0.8369728088378906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.01,0.6520319938659668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.01,0.7440767765045166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.01,0.9737983703613281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.01,1.1946047782897948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.01,1.4376447677612305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,1,power_law_1.01,2.147974395751953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,1,power_law_1.01,4.13267822265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,1,power_law_1.2,0.09880319833755494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,1,power_law_1.2,0.12643840312957763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,1,power_law_1.2,0.15475200414657592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,1,power_law_1.2,0.24462080001831055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,1,power_law_1.2,0.340447998046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,1,power_law_1.2,0.07804160118103028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,1,power_law_1.2,0.42478718757629397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,1,power_law_1.2,0.09658880233764648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,1,power_law_1.2,0.6570112228393554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,1,power_law_1.2,0.11175680160522461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,1,power_law_1.2,0.7177984237670898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,1,power_law_1.2,0.1713152050971985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,1,power_law_1.2,0.7223936080932617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,1,power_law_1.2,0.21541121006011962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,1,power_law_1.2,0.7850111961364746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,1,power_law_1.2,0.2655872106552124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,1,power_law_1.2,0.7872896194458008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,1,power_law_1.2,0.4036223888397217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,1,power_law_1.2,0.800160026550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,1,power_law_1.2,0.4273791790008545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,1,power_law_1.2,0.8526464462280273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,1,power_law_1.2,0.4357952117919922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,1,power_law_1.2,0.44136319160461424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,1,power_law_1.2,0.8712960243225097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,1,power_law_1.2,0.48246397972106936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,1,power_law_1.2,0.8838591575622559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,1,power_law_1.2,0.4955967903137207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,1,power_law_1.2,0.9338560104370117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,1,power_law_1.2,0.5092095851898193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,1,power_law_1.2,0.9186688423156738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,1,power_law_1.2,0.5229695796966553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,1,power_law_1.2,1.024505615234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,1,power_law_1.2,0.5362624168395996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,1,power_law_1.2,1.1072959899902344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,1,power_law_1.2,0.5517119884490966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,1,power_law_1.2,1.2238528251647949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,1,power_law_1.2,0.5903552055358887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,1,power_law_1.2,1.3555328369140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,1,power_law_1.2,0.6495488166809082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,1,power_law_1.2,1.5786944389343263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,1,power_law_1.2,0.7074111938476563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,1,power_law_1.2,1.8592063903808593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,1,power_law_1.2,0.8129280090332032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,1,power_law_1.2,2.3419904708862305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,1,power_law_1.2,0.9100480079650879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,1,power_law_1.2,2.8392704010009764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,1,power_law_1.2,1.1199423789978027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,1,power_law_1.2,3.819910430908203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,1,power_law_1.2,1.319974422454834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,1,power_law_1.2,4.829670333862305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,1,power_law_1.2,1.738105583190918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,1,power_law_1.2,5.754188919067383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,1,power_law_1.2,2.1381248474121093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,1,power_law_1.2,8.67325439453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,1,power_law_1.2,2.96231689453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,1,power_law_1.2,3.8040702819824217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,1,power_law_1.2,16.575949096679686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,1,power_law_1.2,4.626886367797852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,1,power_law_1.2,7.12266845703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,1,power_law_1.2,13.7612548828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,1,power_law_1.2,0.10151040554046631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,1,power_law_1.2,0.1265727996826172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,1,power_law_1.2,0.1315775990486145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,1,power_law_1.2,0.2017280101776123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,1,power_law_1.2,0.2542975902557373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,1,power_law_1.2,0.33194239139556886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,1,power_law_1.2,0.4609536170959473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,1,power_law_1.2,0.4887872219085693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,1,power_law_1.2,0.4970816135406494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,1,power_law_1.2,0.5247168064117431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,1,power_law_1.2,0.5460288047790527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,1,power_law_1.2,0.5407936096191406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,1,power_law_1.2,0.5721983909606934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,1,power_law_1.2,0.5808832168579101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,1,power_law_1.2,0.5948351860046387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,1,power_law_1.2,0.6034560203552246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,1,power_law_1.2,0.022009600698947907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,1,power_law_1.2,0.6302144050598144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,1,power_law_1.2,0.027225598692893982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,1,power_law_1.2,0.7191359996795654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,1,power_law_1.2,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.2,0.7582528114318847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,1,power_law_1.2,0.053395199775695804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.2,0.8763903617858887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,1,power_law_1.2,0.08039680123329163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.2,0.9709568023681641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,1,power_law_1.2,0.11482239961624145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.2,1.197593593597412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,1,power_law_1.2,0.15576319694519042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.2,1.4115455627441407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,1,power_law_1.2,0.16403199434280397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.2,1.8573312759399414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,1,power_law_1.2,0.1756991982460022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.2,2.2840576171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,1,power_law_1.2,0.1818943977355957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.2,3.1564863204956053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,1,power_law_1.2,0.19386240243911743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.2,4.02276496887207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,1,power_law_1.2,0.20160000324249266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,1,power_law_1.2,0.21278080940246583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.2,4.879808044433593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,1,power_law_1.2,0.23675520420074464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,1,power_law_1.2,7.431660461425781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,1,power_law_1.2,0.22652161121368408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,1,power_law_1.2,0.23598079681396483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,1,power_law_1.2,14.363526916503906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,1,power_law_1.2,0.267411208152771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,1,power_law_1.2,0.2699455976486206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.2,0.29596800804138185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.2,0.3084800004959106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.2,0.35548160076141355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.2,0.7508480072021484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.2,0.8780096054077149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.2,0.6483839988708496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.2,0.7585919857025146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.2,1.0089088439941407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.2,1.2178239822387695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.2,1.4697407722473144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,1,power_law_1.2,2.1966848373413086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,1,power_law_1.2,4.231647872924805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,4,balanced,0.036288000643253326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,4,balanced,0.04284266630808512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,4,balanced,0.044682666659355164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,4,balanced,0.06739733119805653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,4,balanced,0.09340799848238628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,4,balanced,0.09703999757766724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,4,balanced,0.09225066502888997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,4,balanced,0.0916319986184438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,4,balanced,0.03585600107908249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,4,balanced,0.09436266620953877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,4,balanced,0.03696533292531967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,4,balanced,0.0410453329483668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,4,balanced,0.09186666210492452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,4,balanced,0.05333333214124044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,4,balanced,0.08894933263460796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,4,balanced,0.06934399902820587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,4,balanced,0.09969066580136617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,4,balanced,0.06952000161012013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,4,balanced,0.09292800227801006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,4,balanced,0.06838933130105336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,4,balanced,0.09384533762931824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,4,balanced,0.06994133194287618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,4,balanced,0.10423466563224792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,4,balanced,0.06993600229422252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,4,balanced,0.06804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,4,balanced,0.0957493285338084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,4,balanced,0.06948266426722209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,4,balanced,0.09977066516876221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,4,balanced,0.07181333502133687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,4,balanced,0.07340799768765767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,4,balanced,0.10152533650398254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,4,balanced,0.07135466734568278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,4,balanced,0.10774399836858113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,4,balanced,0.08044266700744629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,4,balanced,0.07787733276685078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,4,balanced,0.11503466963768005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,4,balanced,0.0795306662718455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,4,balanced,0.08703999718030293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,4,balanced,0.1223520040512085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,4,balanced,0.09302933017412822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,4,balanced,0.10204266508420308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,4,balanced,0.15738133589426676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,4,balanced,0.11652800440788269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,4,balanced,0.17088532447814941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,4,balanced,0.15319466590881348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,4,balanced,0.17659199237823486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,4,balanced,0.22099733352661133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,4,balanced,0.24152000745137533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,4,balanced,0.2718613346417745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,4,balanced,0.30586133400599164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,4,balanced,0.038005332152048744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,4,balanced,0.042805333932240806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,4,balanced,0.04916266600290934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,4,balanced,0.3874346812566121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,4,balanced,0.058821335434913635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,4,balanced,0.08045333127180736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,4,balanced,0.07588266829649608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,4,balanced,0.4437439839045207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,4,balanced,0.07473599910736084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,4,balanced,0.07064533233642578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,4,balanced,0.06927466889222463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,4,balanced,0.0727946658929189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,4,balanced,0.07337066531181335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,4,balanced,0.07453333338101704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,4,balanced,0.0705973356962204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,4,balanced,0.07336533566315968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,4,balanced,0.081194669008255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,4,balanced,0.07912533481915791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,4,balanced,0.08341866731643677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,4,balanced,0.5803306500116984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,4,balanced,0.08561600248018901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,4,balanced,0.5036106506983439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,4,balanced,0.09130133191744487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,4,balanced,0.019472000499566395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,4,balanced,0.022463999688625336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,4,balanced,0.022101332743962605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,4,balanced,0.10225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,4,balanced,0.026159999271233875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,4,balanced,0.03065066784620285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,4,balanced,0.032245332996050514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,4,balanced,0.03474666674931844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,4,balanced,0.036992001036802925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,4,balanced,0.04133866727352142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,4,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,4,balanced,0.04228266576925913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,4,balanced,0.05317866802215576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,4,balanced,0.052144000927607216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,4,balanced,0.7142186959584554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,4,balanced,0.05331199864546458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,4,balanced,0.06389333307743073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,4,balanced,0.06387199958165486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,4,balanced,0.0682239979505539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,4,balanced,0.6149973471959432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,4,balanced,0.08121599753697713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,4,balanced,0.08849599957466125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,4,balanced,0.0904960036277771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,4,balanced,0.10212266445159912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,4,balanced,0.1058026651541392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,4,balanced,0.11382399996121724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,4,balanced,1.0951733589172363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,4,balanced,0.185205340385437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,4,balanced,0.19559999306996664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,4,balanced,0.9321280320485433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,4,balanced,0.27820799748102826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,4,balanced,0.36035199960072833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,4,balanced,0.4418986638387044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,4,balanced,2.1085707346598306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,4,balanced,1.7908159891764324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,4,balanced,0.6937440236409506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,4,balanced,0.11374400059382121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,4,balanced,0.15598932902018228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,4,balanced,0.1789919932683309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,4,balanced,0.24568533897399902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,4,balanced,1.3619573911031086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,4,balanced,0.3126399914423625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,4,balanced,0.44580264886220294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,4,balanced,0.5655786593755087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,4,power_law_1.01,0.04431999921798706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,4,power_law_1.01,0.04039680063724518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,4,power_law_1.01,0.04405759871006012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,4,power_law_1.01,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,4,power_law_1.01,0.06652160286903382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,4,balanced,0.6827733516693115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,4,power_law_1.01,0.08648319840431214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,4,power_law_1.01,0.08710399866104127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,4,power_law_1.01,0.08991360068321227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,4,power_law_1.01,0.0896127998828888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,4,power_law_1.01,0.09056000113487243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,4,power_law_1.01,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,4,power_law_1.01,0.09540479779243469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,4,power_law_1.01,0.09523199796676636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,4,power_law_1.01,0.09459199905395507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,4,power_law_1.01,0.09894400238990783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,4,power_law_1.01,0.09870719909667969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,4,power_law_1.01,0.10398720502853394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,4,power_law_1.01,0.10796799659729003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,4,power_law_1.01,0.12066559791564942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,4,power_law_1.01,0.04391680061817169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,4,power_law_1.01,0.1347839951515198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,4,power_law_1.01,0.04291839897632599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,4,power_law_1.01,0.15135999917984008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,4,power_law_1.01,0.040908798575401306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,4,power_law_1.01,0.18187520503997803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,4,power_law_1.01,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,4,power_law_1.01,0.22615039348602295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,4,balanced,1.0364747047424316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,4,power_law_1.01,0.05581439733505249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,4,power_law_1.01,0.3129791975021362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,4,power_law_1.01,0.06554880142211914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,4,power_law_1.01,0.4022975921630859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,4,power_law_1.01,0.0665727972984314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,4,power_law_1.01,0.5166143894195556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,4,power_law_1.01,0.06750720143318176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,4,power_law_1.01,0.7269375801086426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,4,power_law_1.01,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,4,power_law_1.01,0.7631872177124024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,4,power_law_1.01,0.06883839964866638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,4,power_law_1.01,1.2834943771362304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,4,power_law_1.01,0.06967679858207702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,4,power_law_1.01,0.07260159850120544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,4,power_law_1.01,2.5108800888061524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,4,power_law_1.01,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,4,power_law_1.01,0.07441920042037964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,4,power_law_1.01,0.0775488018989563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,4,power_law_1.01,0.08024960160255432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,4,power_law_1.01,0.08466560244560242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,4,power_law_1.01,0.09327359795570374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,4,power_law_1.01,0.09965440034866332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,4,power_law_1.01,0.12327040433883667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,4,power_law_1.01,0.1444607973098755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,4,power_law_1.01,0.1792512059211731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,4,power_law_1.01,0.21576321125030518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,4,power_law_1.01,0.3237760066986084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,4,power_law_1.01,0.3878335952758789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,4,power_law_1.01,0.5541759967803955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,4,power_law_1.01,0.6826496124267578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,4,power_law_1.01,0.8872256278991699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,4,power_law_1.01,1.4555135726928712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,4,power_law_1.01,3.0398271560668944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,4,balanced,1.985487937927246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,4,power_law_1.01,0.04899199903011322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,4,power_law_1.01,0.0431551992893219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,4,power_law_1.01,0.044486400485038755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,4,power_law_1.01,0.05694720149040222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,4,power_law_1.01,0.05952640175819397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,4,power_law_1.01,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,4,power_law_1.01,0.07309439778327942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,4,power_law_1.01,0.07191680073738098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,4,power_law_1.01,0.07121919989585876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,4,power_law_1.01,0.07439360022544861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,4,power_law_1.01,0.07317759990692138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,4,power_law_1.01,0.07562879920005798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,4,power_law_1.01,0.07377279996871948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,4,power_law_1.01,0.07667199969291687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,4,power_law_1.01,0.08055679798126221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,4,power_law_1.01,0.0840063989162445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,4,power_law_1.01,0.08856319785118102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,4,power_law_1.01,0.10410879850387574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,4,power_law_1.01,0.10936319828033447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,4,power_law_1.01,0.1314303994178772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,4,power_law_1.01,0.14665600061416625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,4,power_law_1.01,0.18531839847564696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,4,power_law_1.01,0.23310720920562744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,4,power_law_1.01,0.30424959659576417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,4,power_law_1.01,0.3678143978118896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,4,power_law_1.01,0.5317823886871338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,4,power_law_1.01,0.7281087875366211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,4,power_law_1.01,0.9532032012939453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,4,power_law_1.01,1.3083647727966308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,4,power_law_1.01,3.019647979736328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,4,power_law_1.01,0.018848000466823576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,4,power_law_1.01,0.019411200284957887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,4,power_law_1.01,0.02282239943742752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,4,power_law_1.01,0.02351360023021698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,4,power_law_1.01,0.02447360008955002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,4,power_law_1.01,0.027372801303863527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,4,power_law_1.01,0.027968001365661622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,4,power_law_1.01,0.02922239899635315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,4,power_law_1.01,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,4,power_law_1.01,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,4,power_law_1.01,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,4,power_law_1.01,0.04227840006351471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,4,power_law_1.01,0.03943040072917938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,4,power_law_1.01,0.04026240110397339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,4,power_law_1.01,0.056543999910354616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,4,power_law_1.01,0.05658239722251892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,4,power_law_1.01,0.059411197900772095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,4,power_law_1.01,0.06453760266304016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,4,power_law_1.01,0.0644864022731781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,4,power_law_1.01,0.07097600102424621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,4,power_law_1.2,0.04695039987564087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,4,power_law_1.01,0.07233920097351074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,4,power_law_1.2,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,4,power_law_1.01,0.09100160002708435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,4,power_law_1.2,0.044409599900245664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,4,power_law_1.2,0.05804160237312317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,4,power_law_1.2,0.0641152024269104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,4,power_law_1.2,0.03891839981079102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,4,power_law_1.2,0.09101439714431762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,4,power_law_1.2,0.03555839955806732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,4,power_law_1.2,0.08906880021095276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,4,power_law_1.2,0.041657599806785586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,4,power_law_1.2,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,4,power_law_1.2,0.04750719964504242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,4,power_law_1.2,0.09252480268478394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,4,power_law_1.2,0.0536575973033905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,4,power_law_1.2,0.09237120151519776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,4,power_law_1.2,0.06466559767723083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,4,power_law_1.2,0.09413759708404541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,4,power_law_1.2,0.06632959842681885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,4,power_law_1.2,0.09477760195732117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,4,power_law_1.2,0.0652288019657135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,4,power_law_1.2,0.09448320269584656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,4,power_law_1.2,0.06807680130004883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,4,power_law_1.2,0.09560319781303406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,4,power_law_1.2,0.06931840181350708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,4,power_law_1.2,0.09996160268783569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,4,power_law_1.2,0.06848000288009644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,4,power_law_1.2,0.10192639827728271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,4,power_law_1.2,0.06707199811935424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,4,power_law_1.2,0.1043776035308838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,4,power_law_1.2,0.07057920098304749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,4,power_law_1.2,0.11525119543075561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,4,power_law_1.2,0.07291520237922669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,4,power_law_1.2,0.078438401222229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,4,power_law_1.2,0.1261888027191162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,4,power_law_1.2,0.08139520287513732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,4,power_law_1.2,0.14206720590591432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,4,power_law_1.2,0.08419839739799499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,4,power_law_1.2,0.16414719820022583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,4,power_law_1.2,0.09447680115699768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,4,power_law_1.2,0.20833919048309327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,4,power_law_1.2,0.10321919918060303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,4,power_law_1.2,0.23526399135589598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,4,power_law_1.2,0.12936320304870605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,4,power_law_1.2,0.33870720863342285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,4,power_law_1.2,0.1446272015571594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,4,power_law_1.2,0.4027008056640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,4,power_law_1.2,0.17645440101623536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,4,power_law_1.2,0.5497983932495117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,4,power_law_1.2,0.22287359237670898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,4,power_law_1.2,0.8469696044921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,4,power_law_1.2,0.332044792175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,4,power_law_1.2,0.9714431762695312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,4,power_law_1.2,0.4226816177368164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,4,power_law_1.2,1.4358016014099122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,4,power_law_1.2,0.648524808883667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,4,power_law_1.2,0.8524352073669433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,4,power_law_1.2,2.6859392166137694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,4,power_law_1.2,1.0502016067504882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,4,power_law_1.2,1.7193279266357422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,4,power_law_1.2,3.7523841857910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,4,power_law_1.2,0.04843519926071167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,4,power_law_1.2,0.04062080085277557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,4,power_law_1.2,0.04240640103816986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,4,power_law_1.2,0.049983999133110045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,4,power_law_1.2,0.0564736008644104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,4,power_law_1.2,0.07230079770088196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,4,power_law_1.2,0.06993280053138733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,4,power_law_1.2,0.073471999168396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,4,power_law_1.2,0.0732096016407013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,4,power_law_1.2,0.07605119943618774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,4,power_law_1.2,0.07339519858360291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,4,power_law_1.2,0.07420799732208253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,4,power_law_1.2,0.07534080147743225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,4,power_law_1.2,0.07992960214614868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,4,power_law_1.2,0.0859391987323761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,4,power_law_1.2,0.08761600255966187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,4,power_law_1.2,0.08999680280685425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,4,power_law_1.2,0.1045375943183899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,4,power_law_1.2,0.11675519943237304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,4,power_law_1.2,0.13214080333709716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,4,power_law_1.2,0.15569920539855958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,4,power_law_1.2,0.20033280849456786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,4,power_law_1.2,0.2412480115890503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,4,power_law_1.2,0.31370880603790285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,4,power_law_1.2,0.4364352226257324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,4,power_law_1.2,0.6537600040435791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,4,power_law_1.2,0.809017562866211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,4,power_law_1.2,1.0203264236450196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,4,power_law_1.2,1.4589183807373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,4,power_law_1.2,3.5600704193115233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,4,power_law_1.01,0.1016767978668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,4,power_law_1.01,0.1328384041786194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,4,power_law_1.01,0.163372802734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,4,power_law_1.01,0.2311552047729492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,4,power_law_1.01,0.2970815896987915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,4,power_law_1.01,0.35097599029541016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,4,power_law_1.01,0.5476607799530029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,4,power_law_1.01,1.0583488464355468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,2,balanced,0.037418665985266365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,4,power_law_1.2,0.018374399840831758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,2,balanced,0.03904533386230469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,4,power_law_1.2,0.01961600035429001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,2,balanced,0.04736533264319102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,4,power_law_1.2,0.020185600221157073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,2,balanced,0.0689386675755183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,4,power_law_1.2,0.021427200734615327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,2,balanced,0.0963253378868103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,2,balanced,0.09264000256856282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,4,power_law_1.2,0.02343039959669113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,2,balanced,0.09038933118184407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,4,power_law_1.2,0.026528000831604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,2,balanced,0.08925333619117737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,4,power_law_1.2,0.028921601176261903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,2,balanced,0.09172800183296204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,4,power_law_1.2,0.02863999903202057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,2,balanced,0.09230400125185649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,4,power_law_1.2,0.031839999556541446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,2,balanced,0.09141866366068523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,4,power_law_1.2,0.03201279938220978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,2,balanced,0.09558932979901631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,2,balanced,0.0934986670811971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,4,power_law_1.2,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,2,balanced,0.09583999713261922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,4,power_law_1.2,0.04046719968318939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,2,balanced,0.1002506713072459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,4,power_law_1.2,0.03916159868240356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,2,balanced,0.09918399651845296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,4,power_law_1.2,0.03656319975852966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,4,power_law_1.2,0.05790079832077026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,2,balanced,0.10500799616177876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,4,power_law_1.2,0.05971840023994446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,2,balanced,0.11000000437100728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,4,power_law_1.2,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,2,balanced,0.1181013286113739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,4,power_law_1.2,0.06907520294189454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,4,power_law_1.2,0.06822400093078614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,2,balanced,0.13192533453305563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,4,power_law_1.2,0.07138559818267823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,4,power_law_1.2,0.07815679907798767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,2,balanced,0.14152000347773233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,4,power_law_1.2,0.09525120258331299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,4,power_law_1.2,0.10913920402526855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,4,power_law_1.2,0.14162559509277345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,2,balanced,0.19157866636912027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,4,power_law_1.2,0.17548160552978515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,4,power_law_1.2,0.2504767894744873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,4,power_law_1.2,0.3008768081665039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,4,power_law_1.2,0.3821248054504395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,2,balanced,0.2188053329785665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,4,power_law_1.2,0.6091263771057129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,4,power_law_1.2,1.1985407829284669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,2,balanced,0.29822399218877155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,2,balanced,0.035946667194366455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,2,balanced,0.3800266583760579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,2,balanced,0.039173332353432976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,2,balanced,0.04381333291530609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,2,balanced,0.05329066514968872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,2,balanced,0.07585600018501282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,2,balanced,0.07255466779073079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,2,balanced,0.07306133210659027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,2,balanced,0.07499200105667114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,2,balanced,0.07532800237337749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,2,balanced,0.07483200232187907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,2,balanced,0.5405386686325073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,2,balanced,0.07533333202203114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,2,balanced,0.07665066421031952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,2,balanced,0.0795306662718455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,2,balanced,0.07978666822115581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,2,balanced,0.08561066786448161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,2,balanced,0.08799999952316284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,2,balanced,0.09170666337013245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,2,balanced,0.104010671377182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,2,balanced,0.11109866698582967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,2,balanced,0.7008586724599203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,2,balanced,0.13329066832860312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,2,balanced,0.14660267035166422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,2,balanced,0.20551466941833496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,2,balanced,0.24411733945210776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,2,balanced,0.8614292939503988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,2,balanced,0.3531573216120402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,2,balanced,0.04437866806983948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,2,balanced,0.0489333321650823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,2,balanced,0.4543466567993164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,2,balanced,0.05649066468079885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,2,balanced,0.07192533214886983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,2,balanced,0.09408000111579895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,2,balanced,0.09039466579755147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,2,balanced,0.08848533034324646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,2,balanced,0.08603733777999878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,2,balanced,0.08506133159001668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,2,balanced,0.08898666501045227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,2,balanced,0.08805867036183675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,2,balanced,0.09195733070373535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,2,balanced,1.328975995381673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,2,balanced,0.08915199836095174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,2,balanced,0.6602986653645834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,2,balanced,0.09459732969601949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,2,balanced,0.10006933410962422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,2,balanced,0.09981866677602132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,2,balanced,0.10566932956377666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,2,balanced,0.11761066317558289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,2,balanced,0.1291039983431498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,2,balanced,0.14966400464375815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,2,balanced,0.17140799760818481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,2,balanced,0.8683306376139323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,2,balanced,0.2434933384259542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,2,balanced,0.29203200340270996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,2,balanced,0.4187306563059489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,2,balanced,1.0801973342895508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,2,balanced,0.5422079960505167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,2,balanced,2.561957359313965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,2,balanced,0.7813333670298258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,2,balanced,0.01942933350801468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,2,balanced,0.02075200031201045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,2,balanced,0.02236266682545344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,2,balanced,0.02607999990383784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,2,balanced,0.030117332935333252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,2,balanced,0.030394665896892548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,2,balanced,0.032298666735490165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,2,balanced,0.032373333970705666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,2,balanced,1.0078026453653972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,2,balanced,0.034602666894594826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,2,balanced,0.03630933413902918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,2,balanced,1.6682133674621582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,2,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,2,balanced,0.040405333042144775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,2,balanced,0.04104000081618627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,2,balanced,0.04204266766707102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,2,balanced,0.05249600112438202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,2,balanced,0.05306133131186167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,2,balanced,0.052853330969810486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,2,balanced,0.07342933118343353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,2,balanced,0.0798826664686203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,2,balanced,1.2395573457082112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,2,balanced,0.09679466485977173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,2,balanced,0.10315733154614766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,2,balanced,0.10383466879526775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,2,balanced,0.11010666688283284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,2,balanced,0.18077333768208823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,2,balanced,1.9234506289164226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,2,balanced,0.18564265966415405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,2,balanced,3.240485191345215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,2,balanced,0.26148800055185956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,2,balanced,0.33552531401316327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,2,balanced,3.697567939758301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,2,balanced,0.4121653238932292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,2,power_law_1.01,0.03963519930839539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,2,power_law_1.01,0.03934719860553741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,2,power_law_1.01,0.044870400428771974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,2,power_law_1.01,0.05512319803237915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,2,power_law_1.01,0.0643455982208252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,2,power_law_1.01,0.0830784022808075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,2,power_law_1.01,0.08409600257873535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,2,power_law_1.01,0.08915200233459472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,2,power_law_1.01,0.08804479837417603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,2,power_law_1.01,0.08929920196533203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,2,power_law_1.01,0.09023360013961793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,2,power_law_1.01,0.09270399808883667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,2,power_law_1.01,0.09466879963874816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,2,power_law_1.01,0.09594879746437072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,2,power_law_1.01,0.10071680545806885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,2,power_law_1.01,0.10324480533599853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,2,power_law_1.01,0.10640640258789062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,2,power_law_1.01,0.11570559740066529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,2,power_law_1.01,0.12663040161132813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,2,power_law_1.01,0.1500864028930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,2,power_law_1.01,0.16652159690856932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,2,balanced,0.6387413342793783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,2,power_law_1.01,0.21025919914245605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,2,power_law_1.01,0.2610752105712891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,2,power_law_1.01,0.35964159965515136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,2,power_law_1.01,0.43386240005493165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,2,power_law_1.01,0.609772777557373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,2,power_law_1.01,0.751193618774414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,2,power_law_1.01,0.9622079849243164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,2,power_law_1.01,1.425171184539795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,2,power_law_1.01,2.7701759338378906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,2,power_law_1.01,0.03607679903507233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,2,power_law_1.01,0.036857599020004274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,2,power_law_1.01,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,2,power_law_1.01,0.04898560047149658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,2,power_law_1.01,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,2,power_law_1.01,0.0691648006439209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,2,power_law_1.01,0.06924160122871399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,2,power_law_1.01,0.0705344021320343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,2,power_law_1.01,0.07355520129203796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,2,power_law_1.01,0.07393280267715455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,2,power_law_1.01,0.07490559816360473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,2,power_law_1.01,0.07834240198135375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,2,power_law_1.01,0.07739520072937012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,2,power_law_1.01,0.08024320006370544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,2,power_law_1.01,0.08537600040435792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,2,power_law_1.01,0.08793600201606751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,2,power_law_1.01,0.09345920085906982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,2,power_law_1.01,0.10552959442138672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,2,balanced,1.246890703837077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,2,power_law_1.01,0.12064640522003174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,2,power_law_1.01,0.14207359552383422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,2,power_law_1.01,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,2,power_law_1.01,0.1718783974647522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,2,power_law_1.01,0.04787200093269348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,2,power_law_1.01,0.22380800247192384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,2,power_law_1.01,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,2,power_law_1.01,0.2874176025390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,2,power_law_1.01,0.06182399988174438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,2,power_law_1.01,0.4137279987335205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,2,power_law_1.01,0.06753280162811279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,2,power_law_1.01,0.08478720188140869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,2,power_law_1.01,0.4862847805023193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,2,power_law_1.01,0.08346239924430847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,2,power_law_1.01,0.7071936130523682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,2,power_law_1.01,0.08699520230293274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,2,power_law_1.01,1.0059519767761231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,2,power_law_1.01,0.08595200181007386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,2,power_law_1.01,1.160524845123291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,2,power_law_1.01,0.08661119937896729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,2,power_law_1.01,1.8117311477661133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,2,power_law_1.01,0.09006720185279846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,2,power_law_1.01,0.09080960154533387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,2,power_law_1.01,3.3811904907226564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,2,power_law_1.01,0.08859519958496094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,2,power_law_1.01,0.09322879910469055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,2,power_law_1.01,0.10021120309829712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,2,power_law_1.01,0.10582400560379028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,2,power_law_1.01,0.11286400556564331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,2,power_law_1.01,0.12423039674758911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,2,power_law_1.01,0.14451199769973755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,2,power_law_1.01,0.17235200405120848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,2,power_law_1.01,0.20213119983673095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,2,power_law_1.01,0.27343358993530276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,2,power_law_1.01,0.3331648111343384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,2,power_law_1.01,0.44954237937927244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,2,power_law_1.01,0.6140223979949951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,2,power_law_1.01,0.837279987335205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,2,power_law_1.01,1.0809087753295898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,2,power_law_1.01,1.2890239715576173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,2,power_law_1.01,2.0703039169311523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,2,power_law_1.01,4.652243041992188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,2,power_law_1.01,0.01706880033016205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,2,power_law_1.01,0.018028800189495087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,2,power_law_1.01,0.019519999623298645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,2,power_law_1.01,0.022009600698947907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,2,power_law_1.01,0.022963200509548188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,2,power_law_1.01,0.02602880001068115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,2,power_law_1.01,0.027238398790359497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,2,power_law_1.01,0.02932479977607727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,2,power_law_1.01,0.030118399858474733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,2,power_law_1.01,0.03030399978160858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,2,power_law_1.01,0.03182719945907593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,2,power_law_1.01,0.03237119913101196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,2,power_law_1.01,0.033983999490737916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,2,power_law_1.01,0.034790399670600894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,2,power_law_1.01,0.04165120124816894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,2,power_law_1.01,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,2,power_law_1.01,0.042131200432777405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,2,power_law_1.01,0.06787840127944947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,2,power_law_1.01,0.07349119782447815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,2,power_law_1.01,0.07675520181655884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,2,power_law_1.01,0.08095999956130981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,2,power_law_1.01,0.09367039799690247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,2,power_law_1.01,0.10886399745941162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,2,power_law_1.01,0.13547519445419312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,2,power_law_1.01,0.1657088041305542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,2,power_law_1.01,0.2246272087097168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,2,power_law_1.01,0.2959552049636841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,2,power_law_1.01,0.34464640617370607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,2,power_law_1.01,0.5311999797821045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,2,power_law_1.01,1.0546624183654785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,2,power_law_1.2,0.040243199467658995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,2,power_law_1.2,0.04073599874973297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,2,power_law_1.2,0.04437119960784912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,2,power_law_1.2,0.05137280225753784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,2,power_law_1.2,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,2,power_law_1.2,0.08539519906044006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,2,power_law_1.2,0.08938239812850952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,2,power_law_1.2,0.09114879965782166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,2,power_law_1.2,0.08972160220146179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,2,power_law_1.2,0.0901311993598938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,2,power_law_1.2,0.09257599711418152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,2,power_law_1.2,0.09367039799690247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,2,power_law_1.2,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,2,power_law_1.2,0.09477760195732117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,2,power_law_1.2,0.03641600012779236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,2,power_law_1.2,0.0982912003993988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,2,power_law_1.2,0.039078399538993835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,2,power_law_1.2,0.10140800476074219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,2,power_law_1.2,0.04535680115222931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,2,power_law_1.2,0.10220799446105958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,2,power_law_1.2,0.052883201837539674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,2,power_law_1.2,0.10607999563217163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,2,power_law_1.2,0.06933119893074036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,2,power_law_1.2,0.1229248046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,2,power_law_1.2,0.06983680129051209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,2,power_law_1.2,0.1302016019821167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,2,power_law_1.2,0.07221119999885559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,2,power_law_1.2,0.14769920110702514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,2,power_law_1.2,0.07367680072784424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,2,power_law_1.2,0.1710911989212036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,2,power_law_1.2,0.07384960055351257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,2,power_law_1.2,0.21897599697113038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,2,power_law_1.2,0.07587839961051941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,2,power_law_1.2,0.07735679745674133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,2,power_law_1.2,0.2637824058532715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,2,power_law_1.2,0.077183997631073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,2,power_law_1.2,0.35533440113067627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,2,power_law_1.2,0.08131200075149536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,2,power_law_1.2,0.45150079727172854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,2,power_law_1.2,0.08711040019989014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,2,power_law_1.2,0.7006144046783447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,2,power_law_1.2,0.08826239705085755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,2,power_law_1.2,0.8963711738586426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,2,power_law_1.2,0.09292160272598267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,2,power_law_1.2,1.0495295524597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,2,power_law_1.2,0.046982398629188536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,2,power_law_1.2,0.10864640474319458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,2,power_law_1.2,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,2,power_law_1.2,1.5229120254516602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,2,power_law_1.2,0.12271360158920289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,2,power_law_1.2,0.04933120012283325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,2,power_law_1.2,0.14855680465698243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,2,power_law_1.2,3.3071807861328124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,2,power_law_1.2,0.057708799839019775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,2,power_law_1.2,0.17393280267715455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,2,power_law_1.2,0.06479359865188598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,2,power_law_1.2,0.2274240016937256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,2,power_law_1.2,0.08479359745979309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,2,power_law_1.2,0.299564790725708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,2,power_law_1.2,0.08236160278320312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,2,power_law_1.2,0.40166401863098145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,2,power_law_1.2,0.08822399973869324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,2,power_law_1.2,0.5286528110504151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,2,power_law_1.2,0.08700799942016602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,2,power_law_1.2,0.08818560242652893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,2,power_law_1.2,0.8205375671386719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,2,power_law_1.2,0.08881279826164246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,2,power_law_1.2,1.0452672004699708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,2,power_law_1.2,0.09188479781150818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,2,power_law_1.2,1.2279232025146485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,2,power_law_1.2,0.09379199743270875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,2,power_law_1.2,1.9189695358276366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,2,power_law_1.2,0.09525120258331299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,2,power_law_1.2,0.10109440088272095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,2,power_law_1.2,3.9024063110351563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,2,power_law_1.2,0.10464639663696289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,2,power_law_1.2,0.11500799655914307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,2,power_law_1.2,0.12906880378723146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,2,power_law_1.2,0.14897279739379882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,2,power_law_1.2,0.17434240579605104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,2,power_law_1.2,0.20426878929138184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,2,power_law_1.2,0.2646336078643799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,2,power_law_1.2,0.34673919677734377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,2,power_law_1.2,0.489350414276123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,2,power_law_1.2,0.6205120086669922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,2,power_law_1.2,0.9173055648803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,2,power_law_1.2,1.147532844543457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,2,power_law_1.2,1.4838015556335449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,2,power_law_1.2,2.3460159301757812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,2,power_law_1.2,4.553388977050782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,2,power_law_1.2,0.01712000072002411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,2,power_law_1.2,0.018355199694633485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,2,power_law_1.2,0.019788800179958342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,2,power_law_1.2,0.020908799767494202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,2,power_law_1.2,0.022387200593948366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,2,power_law_1.2,0.025209599733352663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,2,power_law_1.2,0.027008000016212463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,2,power_law_1.2,0.02924799919128418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,2,power_law_1.2,0.030636799335479737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,2,power_law_1.2,0.03113600015640259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,2,power_law_1.2,0.0321152001619339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,2,power_law_1.2,0.0324288010597229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,2,power_law_1.2,0.032025599479675294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,2,power_law_1.2,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,2,power_law_1.2,0.040531200170516965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,2,power_law_1.2,0.041126400232315063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,2,power_law_1.2,0.0417279988527298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,2,power_law_1.2,0.06915199756622314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,2,power_law_1.2,0.07624959945678711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,2,power_law_1.2,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,2,power_law_1.2,0.08376960158348083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,2,power_law_1.2,0.09662719964981079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,2,power_law_1.2,0.1096384048461914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,2,power_law_1.2,0.14088959693908693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,2,power_law_1.2,0.17297919988632202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,2,power_law_1.2,0.2375744104385376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,2,power_law_1.2,0.2981503963470459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,2,power_law_1.2,0.36978559494018554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,2,power_law_1.2,0.5138175964355469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,2,power_law_1.2,1.0704575538635255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,1,balanced,0.04223999877770742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,1,balanced,0.04493333399295807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,1,balanced,0.05216533442338308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,1,balanced,0.07481599847475688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,1,balanced,0.10771733522415161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,1,balanced,0.10335999727249146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,1,balanced,0.10355200370152791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,1,balanced,0.10475732882817586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,1,balanced,0.10458667079607646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,1,balanced,0.10407466689745586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,1,balanced,0.10498666763305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,1,balanced,0.03913066784540812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,1,balanced,0.10973866780598958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,1,balanced,0.04438933233420054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,1,balanced,0.048165331284205117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,1,balanced,0.10829866925875346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,1,balanced,0.06343466540177663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,1,balanced,0.11177600423494975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,1,balanced,0.09038933118184407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,1,balanced,0.11736533045768738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,1,balanced,0.08851200342178345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,1,balanced,0.08804266651471455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,1,balanced,0.11744532982508342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,1,balanced,0.0902400016784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,1,balanced,0.124399999777476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,1,balanced,0.09122666716575623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,1,balanced,0.0916266640027364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,1,balanced,0.1379093329111735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,1,balanced,0.09342933694521587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,1,balanced,0.09524266918500264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,1,balanced,0.1495253344376882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,1,balanced,0.0990613301595052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,1,balanced,0.1011253297328949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,1,power_law_1.01,0.03694080114364624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,1,balanced,0.17081600427627563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,1,balanced,0.10617599884668986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,1,power_law_1.01,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,1,power_law_1.01,0.04750719964504242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,1,balanced,0.11162666479746501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,1,balanced,0.19364800055821738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,1,power_law_1.01,0.05653759837150574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,1,balanced,0.11788266897201538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,1,power_law_1.01,0.07015039920806884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,1,balanced,0.13750400145848593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,1,power_law_1.01,0.09042559862136841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,1,power_law_1.01,0.09441919922828675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,1,balanced,0.28149867057800293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,1,balanced,0.14967466394106546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,1,power_law_1.01,0.09656320214271545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,1,power_law_1.01,0.09708160161972046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,1,balanced,0.18452799320220947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,1,power_law_1.01,0.10005760192871094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,1,power_law_1.01,0.10100480318069457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,1,balanced,0.21207465728123984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,1,balanced,0.3304799993832906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,1,power_law_1.01,0.1037824034690857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,1,power_law_1.01,0.10663039684295654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,1,power_law_1.01,0.10906879901885987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,1,balanced,0.3235519925753276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,1,power_law_1.01,0.11644799709320068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,1,power_law_1.01,0.11635199785232545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,1,power_law_1.01,0.12216960191726685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,1,balanced,0.4617760181427002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,1,power_law_1.01,0.13973759412765502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,1,balanced,0.3940906524658203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,1,power_law_1.01,0.1540544033050537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,1,power_law_1.01,0.1835584044456482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,1,power_law_1.01,0.2151360034942627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,1,power_law_1.01,0.2827647924423218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,1,balanced,0.5708693265914917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,1,power_law_1.01,0.34213759899139407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,1,power_law_1.01,0.4765120029449463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,1,balanced,0.593290646870931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,1,power_law_1.01,0.5952640056610108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,1,power_law_1.01,0.8425215721130371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,1,power_law_1.01,1.088595199584961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,1,balanced,0.7509653568267822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,1,power_law_1.01,1.3371968269348145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,1,power_law_1.01,2.0777984619140626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,1,power_law_1.01,4.087955093383789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,1,balanced,0.8537333011627197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,1,balanced,1.1044480005900066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,1,balanced,1.4611199696858723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,1,balanced,1.114693323771159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,1,balanced,1.8184107144673665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,1,power_law_1.01,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,1,power_law_1.01,0.039699199795722964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,1,power_law_1.01,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,1,power_law_1.01,0.055327999591827395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,1,power_law_1.01,0.06059520244598389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,1,balanced,1.3749440511067708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,1,power_law_1.01,0.07916160225868225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,1,power_law_1.01,0.0807039976119995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,1,power_law_1.01,0.08377599716186523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,1,power_law_1.01,0.08293759822845459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,1,power_law_1.01,0.08673279881477355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,1,power_law_1.01,0.08809599876403809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,1,power_law_1.01,0.091839998960495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,1,power_law_1.01,0.09466879963874816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,1,power_law_1.01,0.09699199795722961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,1,power_law_1.01,0.10410239696502685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,1,power_law_1.01,0.10776959657669068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,1,power_law_1.01,0.1164736032485962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,1,power_law_1.01,0.1378559947013855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,1,power_law_1.01,0.15797120332717896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,1,balanced,2.860741297403971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,1,power_law_1.01,0.19621119499206544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,1,power_law_1.01,0.23877758979797364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,1,power_law_1.01,0.32604160308837893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,1,power_law_1.01,0.4123072147369385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,1,power_law_1.01,0.585862398147583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,1,power_law_1.01,0.7562047958374023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,1,power_law_1.01,1.1042816162109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,1,power_law_1.01,1.4443327903747558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,1,power_law_1.01,1.7911231994628907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,1,power_law_1.01,2.8101055145263674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,1,power_law_1.01,5.537350463867187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,1,balanced,2.1205652554829917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,1,power_law_1.2,0.037862399220466615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,1,power_law_1.2,0.04195840060710907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,1,balanced,5.596448262532552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,1,power_law_1.2,0.04689919948577881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,1,power_law_1.2,0.05426560044288635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,1,power_law_1.2,0.06661760210990905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,1,power_law_1.2,0.0906495988368988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,1,power_law_1.2,0.09668480157852173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,1,power_law_1.2,0.09748479723930359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,1,power_law_1.2,0.09919360280036926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,1,power_law_1.2,0.10288000106811523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,1,power_law_1.2,0.10133119821548461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,1,power_law_1.2,0.10512000322341919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,1,power_law_1.2,0.106278395652771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,1,power_law_1.2,0.11052800416946411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,1,power_law_1.2,0.11598720550537109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,1,power_law_1.2,0.11768319606781005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,1,power_law_1.2,0.12479360103607177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,1,power_law_1.2,0.1425279974937439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,1,power_law_1.2,0.15717120170593263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,1,power_law_1.2,0.18606079816818238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,1,power_law_1.2,0.21593599319458007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,1,power_law_1.2,0.2828160047531128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,1,power_law_1.2,0.34306559562683103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,1,power_law_1.2,0.4759488105773926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,1,balanced,4.106810569763184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,1,power_law_1.2,0.5896192073822022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,1,power_law_1.2,0.8392319679260254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,1,power_law_1.2,1.0836159706115722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,1,power_law_1.2,1.3371328353881835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,1,power_law_1.2,2.0820735931396483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,1,power_law_1.2,4.080966567993164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,1,power_law_1.2,0.03646079897880554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,1,power_law_1.2,0.04015359878540039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,1,power_law_1.2,0.04344319999217987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,1,power_law_1.2,0.05367680191993714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,1,power_law_1.2,0.05798400044441223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,1,power_law_1.2,0.07723519802093506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,1,power_law_1.2,0.0816703975200653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,1,power_law_1.2,0.08389760255813598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,1,power_law_1.2,0.08650879859924317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,1,power_law_1.2,0.08726400136947632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,1,power_law_1.2,0.09055359959602356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,1,power_law_1.2,0.09326720237731934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,1,power_law_1.2,0.0945088028907776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,1,power_law_1.2,0.09651839733123779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,1,power_law_1.2,0.10536320209503174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,1,power_law_1.2,0.10866559743881225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,1,power_law_1.2,0.11880960464477539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,1,power_law_1.2,0.14008320569992067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,1,power_law_1.2,0.15975040197372437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,1,power_law_1.2,0.19896320104599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,1,power_law_1.2,0.23800959587097167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,1,power_law_1.2,0.32617599964141847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,1,power_law_1.2,0.4095104217529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,1,power_law_1.2,0.5864831924438476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,1,power_law_1.2,0.7594624042510987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,1,power_law_1.2,1.0992959976196288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,1,power_law_1.2,1.442521572113037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,1,power_law_1.2,1.7836288452148437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,1,power_law_1.2,2.8051328659057617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,4,balanced,0.056559999783833824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,4,balanced,0.07374933362007141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,1,power_law_1.2,5.524601745605469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,4,balanced,0.10439999898274739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,4,balanced,0.1579146683216095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,4,balanced,0.2645333409309387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,4,balanced,0.2698240081469218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,4,balanced,0.28726400931676227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,4,balanced,0.2691466609636943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,4,balanced,0.2744266589482625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,4,balanced,0.27321600914001465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,4,balanced,0.25867199897766113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,4,balanced,0.28350400924682617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,4,balanced,0.26018667221069336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,4,balanced,0.26577599843343097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,4,balanced,0.27086933453877765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,4,balanced,0.26154132684071857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,4,balanced,0.2609600027402242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,4,balanced,0.2579519947369893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,4,balanced,0.26071999470392865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,4,balanced,0.04650666813055674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,4,balanced,0.05261866748332977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,4,balanced,0.2736213405927022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,4,balanced,0.07259200016657512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,4,balanced,0.05226133267084757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,4,balanced,0.1014400025208791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,4,balanced,0.05560533205668131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,4,balanced,0.2860426704088847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,4,balanced,0.15989333391189575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,4,balanced,0.07966400186220805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,4,balanced,0.12062933047612508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,4,balanced,0.17061867316563925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,4,balanced,0.18503999710083008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,4,balanced,0.15156267086664835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,4,balanced,0.16744534174601236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,4,balanced,0.15986667076746622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,4,balanced,0.4206613302230835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,4,balanced,0.16172266999880472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,4,balanced,0.1606666644414266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,4,balanced,0.16471999883651733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,4,balanced,0.15424000223477682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,4,balanced,0.1590773363908132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,4,balanced,0.14750933647155762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,4,balanced,0.1622666617234548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,4,balanced,0.16150933504104614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,4,balanced,0.4036639928817749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,4,balanced,0.14984533190727234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,4,balanced,0.15955199797948202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,4,balanced,0.16039466857910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,4,balanced,0.15990400314331055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,4,balanced,0.15899733702341715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,4,balanced,0.17938133080800375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,4,balanced,0.14594133694966635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,4,balanced,0.16038933396339417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,4,balanced,0.17535465955734253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,4,balanced,0.5349653164545695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,4,balanced,0.1716960072517395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,4,balanced,0.15810666481653848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,4,balanced,0.17162134250005087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,4,balanced,0.15843199690183005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,4,balanced,0.15873600045839945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,4,balanced,0.1792746583620707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,4,balanced,0.17297067244847616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,4,balanced,0.20109333594640097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,4,balanced,0.6675573190053304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,4,balanced,0.17433067162831625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,4,balanced,0.218176007270813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,4,balanced,0.19326400756835938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,4,balanced,0.2965493400891622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,4,balanced,0.298309326171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,4,balanced,0.3181493282318115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,4,balanced,0.30904533465703327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,4,balanced,0.9558986822764078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,4,balanced,0.4256853262583415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,4,balanced,0.43930665651957196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,4,balanced,0.5261013507843018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,4,balanced,1.2406453291575115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,4,balanced,0.5576853354771932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,4,balanced,0.7386186917622884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,4,balanced,0.8013226985931396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,4,balanced,1.535754680633545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,4,balanced,0.9673493703206381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,4,balanced,1.0372426509857178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,4,balanced,0.025498665869235992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,4,balanced,0.029733332494894665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,4,balanced,1.1886506875356038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,4,balanced,0.03664533297220866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,4,balanced,0.05065066615740458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,4,balanced,0.07125866909821828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,4,balanced,0.07437866429487865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,4,balanced,0.07693333427111308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,4,balanced,0.07986666758855183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,4,balanced,0.09964799880981445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,4,balanced,2.4052000045776367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,4,balanced,0.09419199824333191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,4,balanced,0.09742400050163269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,4,balanced,0.14667200048764548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,4,balanced,0.12939199805259705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,4,balanced,1.2774346669514973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,4,balanced,0.13706666231155396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,4,balanced,0.15718932946523032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,4,balanced,0.1606986622015635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,4,balanced,0.17749865849812826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,4,balanced,0.1904533306757609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,4,balanced,0.19926933447519937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,4,balanced,0.21250667174657187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,4,balanced,1.8281386693318684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,4,balanced,0.2239946722984314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,4,balanced,0.23368000984191895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,4,balanced,0.25277866919835407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,4,balanced,0.41868801911671955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,4,balanced,1.9793012936909993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,4,balanced,0.4450773398081462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,4,balanced,0.6415520111719767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,4,balanced,4.799754778544108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,4,balanced,3.5822347005208335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,4,balanced,0.8459893067677816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,4,balanced,1.0520106951395671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,4,balanced,3.89028263092041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,4,balanced,1.659546693166097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,4,balanced,3.280127843221029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,4,power_law_1.01,0.10540159940719604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,4,power_law_1.01,0.06661120057106018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,4,power_law_1.01,0.09489920139312744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,4,power_law_1.01,0.1267135977745056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,4,power_law_1.01,0.16241919994354248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,4,power_law_1.01,0.23352320194244386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,4,power_law_1.01,0.24657280445098878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,4,power_law_1.01,0.07729920148849487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,4,power_law_1.01,0.24825599193572997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,4,power_law_1.01,0.06561279892921448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,4,power_law_1.01,0.2544447898864746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,4,power_law_1.01,0.06627200245857238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,4,power_law_1.01,0.24868481159210204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,4,power_law_1.01,0.10109440088272095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,4,power_law_1.01,0.2622335910797119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,4,power_law_1.01,0.10891519784927368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,4,power_law_1.01,0.26462080478668215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,4,power_law_1.01,0.14012160301208496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,4,power_law_1.01,0.2475071907043457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,4,power_law_1.01,0.14793599843978883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,4,power_law_1.01,0.2722559928894043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,4,power_law_1.01,0.15523200035095214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,4,power_law_1.01,0.24925439357757567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,4,power_law_1.01,0.2648576021194458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,4,power_law_1.01,0.15624959468841554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,4,power_law_1.01,0.27204480171203616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,4,power_law_1.01,0.15368959903717042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,4,power_law_1.01,0.3027008056640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,4,power_law_1.01,0.15581439733505248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,4,power_law_1.01,0.3253056049346924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,4,power_law_1.01,0.16045440435409547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,4,power_law_1.01,0.37695999145507814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,4,power_law_1.01,0.15953279733657838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,4,power_law_1.01,0.411359977722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,4,power_law_1.01,0.1510655999183655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,4,power_law_1.01,0.5060416221618652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,4,power_law_1.01,0.16471680402755737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,4,power_law_1.01,0.16428159475326537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,4,power_law_1.01,0.6476672172546387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,4,power_law_1.01,0.17777279615402222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,4,power_law_1.01,0.8038528442382813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,4,power_law_1.01,0.19615360498428344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,4,power_law_1.01,1.061689567565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,4,power_law_1.01,0.2184000015258789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,4,power_law_1.01,0.07523199915885925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,4,power_law_1.01,1.332364845275879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,4,power_law_1.01,0.25752959251403806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,4,power_law_1.01,0.06650879979133606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,4,power_law_1.01,1.9083520889282226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,4,power_law_1.01,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,4,power_law_1.01,0.2980544090270996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,4,power_law_1.01,2.4062335968017576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,4,power_law_1.01,0.10449919700622559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,4,power_law_1.01,0.3749056100845337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,4,power_law_1.01,3.3556480407714844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,4,power_law_1.01,0.10344959497451782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,4,power_law_1.01,0.47706241607666017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,4,power_law_1.01,0.14652800559997559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,4,power_law_1.01,0.6241087913513184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,4,power_law_1.01,6.703469085693359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,4,power_law_1.01,0.15714559555053711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,4,power_law_1.01,0.8063488006591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,4,power_law_1.01,0.14512640237808228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,4,power_law_1.01,1.023027229309082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,4,power_law_1.01,0.15965440273284912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,4,power_law_1.01,1.3960000038146974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,4,power_law_1.01,0.16049280166625976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,4,power_law_1.01,1.6032255172729493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,4,power_law_1.01,0.16225279569625856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,4,power_law_1.01,0.1665984034538269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,4,power_law_1.01,2.6549823760986326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,4,power_law_1.01,0.16344319581985473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,4,power_law_1.01,5.192287826538086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,4,power_law_1.01,0.16185599565505981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,4,power_law_1.01,0.16828160285949706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,4,power_law_1.01,0.1730687975883484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,4,power_law_1.01,0.18050559759140014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,4,power_law_1.01,0.19370239973068237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.01,0.21233279705047609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.01,0.25858559608459475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.01,0.30889599323272704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.01,0.38551039695739747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.01,0.4600192070007324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.01,0.5982912063598633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.01,0.7194240093231201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.01,1.073209571838379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.01,1.285267162322998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,4,power_law_1.01,1.4631104469299316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,4,power_law_1.01,2.3687616348266602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,4,power_law_1.01,4.91712646484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,4,power_law_1.01,0.026732799410820008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,4,power_law_1.01,0.025862398743629455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,4,power_law_1.01,0.03308799862861633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,4,power_law_1.01,0.04319359958171844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,4,power_law_1.01,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,4,power_law_1.01,0.06474879980087281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,4,power_law_1.01,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,4,power_law_1.01,0.06738560199737549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,4,power_law_1.01,0.07731199860572815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,4,power_law_1.01,0.07344639897346497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,4,power_law_1.01,0.07111039757728577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,4,power_law_1.01,0.10199040174484253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,4,power_law_1.01,0.08858240246772767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,4,power_law_1.01,0.08540160059928895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,4,power_law_1.01,0.1334272027015686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,4,power_law_1.01,0.1353983998298645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,4,power_law_1.01,0.1478656053543091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,4,power_law_1.01,0.13967360258102418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.01,0.13644800186157227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.01,0.14464000463485718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.01,0.1509951949119568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.01,0.18337279558181763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.01,0.22466559410095216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.01,0.29134719371795653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.01,0.3758399963378906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.01,0.5302080154418946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.01,0.6558464050292969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,4,power_law_1.01,0.8690431594848633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,4,power_law_1.01,1.2252479553222657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,4,power_law_1.01,2.543699264526367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,4,power_law_1.2,0.10404479503631592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,4,power_law_1.2,0.06999040246009827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,4,power_law_1.2,0.09763200283050537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,4,power_law_1.2,0.12799359560012818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,4,power_law_1.2,0.15667840242385864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,4,power_law_1.2,0.2326591968536377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,4,power_law_1.2,0.2588864088058472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,4,power_law_1.2,0.2588799953460693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,4,power_law_1.2,0.24213759899139403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,4,power_law_1.2,0.255513596534729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,4,power_law_1.2,0.2560960054397583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,4,power_law_1.2,0.07403519749641418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,4,power_law_1.2,0.27479679584503175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,4,power_law_1.2,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,4,power_law_1.2,0.27093119621276857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,4,power_law_1.2,0.061689597368240354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,4,power_law_1.2,0.2674623966217041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,4,power_law_1.2,0.07675520181655884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,4,power_law_1.2,0.27983999252319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,4,power_law_1.2,0.0970304012298584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,4,power_law_1.2,0.28248960971832277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,4,power_law_1.2,0.14476799964904785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,4,power_law_1.2,0.2855168104171753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,4,power_law_1.2,0.1477120041847229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,4,power_law_1.2,0.31021440029144287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,4,power_law_1.2,0.14886399507522582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,4,power_law_1.2,0.3597887992858887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,4,power_law_1.2,0.15404800176620484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,4,power_law_1.2,0.4065728187561035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,4,power_law_1.2,0.15166720151901245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,4,power_law_1.2,0.47047038078308107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,4,power_law_1.2,0.15822720527648926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,4,power_law_1.2,0.5536960124969482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,4,power_law_1.2,0.15804799795150756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,4,power_law_1.2,0.7142591953277588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,4,power_law_1.2,0.1667904019355774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,4,power_law_1.2,0.8917823791503906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,4,power_law_1.2,0.1673535943031311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,4,power_law_1.2,0.9746687889099122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,4,power_law_1.2,0.17244160175323486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,4,power_law_1.2,1.5941503524780274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,4,power_law_1.2,0.17848960161209107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,4,power_law_1.2,0.08283519744873047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,4,power_law_1.2,2.171392059326172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,4,power_law_1.2,0.18085119724273682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,4,power_law_1.2,0.0580672025680542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,4,power_law_1.2,2.566783905029297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,4,power_law_1.2,0.20103681087493896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,4,power_law_1.2,0.07346559762954712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,4,power_law_1.2,0.23006720542907716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,4,power_law_1.2,0.09370239973068237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,4,power_law_1.2,4.07586555480957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,4,power_law_1.2,0.11742719411849975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,4,power_law_1.2,0.2854975938796997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,4,power_law_1.2,0.1518336057662964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,4,power_law_1.2,0.3075263977050781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,4,power_law_1.2,8.944505310058593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,4,power_law_1.2,0.15208959579467773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,4,power_law_1.2,0.3842495918273926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,4,power_law_1.2,0.16264959573745727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,4,power_law_1.2,0.4944896221160889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,4,power_law_1.2,0.16183040142059327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,4,power_law_1.2,0.6466368198394775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,4,power_law_1.2,0.166975998878479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,4,power_law_1.2,0.8529472351074219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,4,power_law_1.2,0.16305279731750488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,4,power_law_1.2,1.219968032836914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,4,power_law_1.2,0.17180800437927246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,4,power_law_1.2,1.4827775955200195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,4,power_law_1.2,0.1633471965789795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,4,power_law_1.2,1.8894784927368165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,4,power_law_1.2,0.1667072057723999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,4,power_law_1.2,0.17052160501480101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,4,power_law_1.2,3.355052947998047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,4,power_law_1.2,0.1818176031112671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,4,power_law_1.2,6.561555480957031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,4,power_law_1.2,0.18870400190353392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,4,power_law_1.2,0.21130878925323487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.2,0.22371840476989746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.2,0.2671168088912964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.2,0.3020031929016113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.2,0.3877311944961548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.2,0.45235838890075686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.2,0.6271488189697265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.2,0.8154303550720214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.2,1.0863295555114747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.2,1.415609645843506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,4,power_law_1.2,1.678054428100586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,4,power_law_1.2,2.914784049987793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,4,power_law_1.2,5.581158447265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,4,power_law_1.2,0.026726400852203368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,4,power_law_1.2,0.027084800601005554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,4,power_law_1.2,0.03425279855728149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,4,power_law_1.2,0.041043201088905336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,4,power_law_1.2,0.04429439902305603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,4,power_law_1.2,0.06345599889755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,4,power_law_1.2,0.0676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,4,power_law_1.2,0.06933760046958923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,4,power_law_1.2,0.07572479844093323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,4,power_law_1.2,0.07434880137443542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,4,power_law_1.2,0.07415680289268493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,4,power_law_1.2,0.09966080188751221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,4,power_law_1.2,0.08910080194473266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,4,power_law_1.2,0.09036160111427308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,4,power_law_1.2,0.13956480026245116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,4,power_law_1.2,0.13598719835281373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,4,power_law_1.2,0.15355520248413085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,4,power_law_1.2,0.13288320302963258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.2,0.14382719993591309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.2,0.15400320291519165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.2,0.1667072057723999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.2,0.19916800260543824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.2,0.24440319538116456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.2,0.3221951961517334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.2,0.38147199153900146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.2,0.5288000106811523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.2,0.6668992042541504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,4,power_law_1.2,0.8277824401855469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,4,power_law_1.2,1.3636672019958496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,4,power_law_1.2,2.7715904235839846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,2,balanced,0.051029334465662636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,2,balanced,0.06619200110435486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,2,balanced,0.09912533561388652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,2,balanced,0.15704533457756042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,2,balanced,0.2632426619529724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,2,balanced,0.26344533761342365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,2,balanced,0.25737067063649494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,2,balanced,0.24206932385762533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,2,balanced,0.27135999997456867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,2,balanced,0.2564106583595276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,2,balanced,0.04270400106906891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,2,balanced,0.250874658425649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,2,balanced,0.05287466446558634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,2,balanced,0.2669493357340495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,2,balanced,0.07235200206438701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,2,balanced,0.2502239942550659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,2,balanced,0.1030613382657369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,2,balanced,0.15945600469907126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,2,balanced,0.25734933217366535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,2,balanced,0.16371200482050577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,2,balanced,0.2598186731338501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,2,balanced,0.152730663617452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,2,balanced,0.2553013364473979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,2,balanced,0.15837333599726358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,2,balanced,0.1613706648349762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,2,balanced,0.2637866735458374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,2,balanced,0.15747732917467752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,2,balanced,0.27106666564941406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,2,balanced,0.1527359982331594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,2,balanced,0.16553599635759988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,2,balanced,0.2818400065104167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,2,balanced,0.15614400307337442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,2,balanced,0.16179733475049338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,2,balanced,0.30267200867335003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,2,balanced,0.17363200585047403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,2,balanced,0.16833066940307617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,2,balanced,0.05454933146635691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,2,balanced,0.32468267281850177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,2,balanced,0.1777226726214091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,2,balanced,0.0621013343334198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,2,balanced,0.08037333190441132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,2,balanced,0.19197332859039307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,2,balanced,0.11712533235549927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,2,balanced,0.17679466803868613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,2,balanced,0.45791999499003094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,2,balanced,0.20851733287175497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,2,balanced,0.17200533548990884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,2,balanced,0.17152533928553262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,2,balanced,0.1723946730295817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,2,balanced,0.23545600970586142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,2,balanced,0.165583997964859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,2,balanced,0.1734559933344523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,2,balanced,0.47999465465545654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,2,balanced,0.1637493371963501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,2,balanced,0.26480533679326373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,2,balanced,0.16524266203244528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,2,balanced,0.17519466082255045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,2,balanced,0.16126400232315063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,2,balanced,0.3662079970041911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,2,balanced,0.18428266048431396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,2,balanced,0.17730667193730673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,2,balanced,0.6398559808731079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,2,balanced,0.17785600821177164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,2,balanced,0.42299731572469074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,2,balanced,0.19033600886662802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,2,balanced,0.20458134015401205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,2,balanced,0.22245866060256958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,2,balanced,0.8069547017415365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,2,balanced,0.5862400134404501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,2,balanced,0.25075199206670123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,2,balanced,0.386186679204305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,2,balanced,0.7522506713867188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,2,balanced,0.4293760061264038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,2,balanced,1.16977063814799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,2,balanced,0.5994879802068075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,2,balanced,1.085909366607666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,2,balanced,0.7574079831441244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,2,balanced,1.5261227289835613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,2,balanced,1.4165333112080891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,2,balanced,0.023786666492621105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,2,balanced,1.0795306364695232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,2,balanced,0.0262773334980011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,2,balanced,0.032560000816980995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,2,balanced,0.0491893341143926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,2,balanced,0.07160000006357829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,2,balanced,0.07339199880758922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,2,balanced,0.07740266621112823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,2,balanced,0.08003200093905131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,2,balanced,0.07980800171693166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,2,balanced,0.08083733419577281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,2,balanced,0.08309333523114522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,2,balanced,0.09670933087666829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,2,balanced,1.7491307258605957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,2,balanced,0.09585066636403401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,2,balanced,1.3932372728983562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,2,balanced,0.09337600072224934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,2,balanced,0.13715733091036478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,2,balanced,1.8930667241414387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,2,balanced,0.13061867157618204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,2,balanced,0.12416000167528789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,2,balanced,0.17693867286046347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,2,balanced,0.19653334220250449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,2,balanced,0.20703999201456705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,2,balanced,0.21322667598724365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,2,balanced,1.7095840771993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,2,balanced,0.208624005317688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,2,balanced,0.21497599283854166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,2,balanced,2.7452481587727866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,2,balanced,0.368613322575887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,2,balanced,0.36611199378967285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,2,balanced,2.992783864339193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,2,balanced,2.657855987548828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,2,balanced,0.5302079916000366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,2,balanced,0.6919626394907633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,2,balanced,5.400362650553386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,2,balanced,0.856607993443807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,2,balanced,6.038506825764974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,2,balanced,5.159194628397624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,2,balanced,1.3635892868041992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,2,balanced,2.673386573791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,2,power_law_1.01,0.06776959896087646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,2,power_law_1.01,0.06476160287857055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,2,power_law_1.01,0.07700480222702026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,2,power_law_1.01,0.12169599533081055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,2,power_law_1.01,0.15238399505615235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,2,power_law_1.01,0.21738879680633544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,2,power_law_1.01,0.23802239894866944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,2,power_law_1.01,0.2427903890609741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,2,power_law_1.01,0.23413119316101075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,2,power_law_1.01,0.2395711898803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,2,power_law_1.01,0.2435136079788208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,2,power_law_1.01,0.24757759571075438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,2,power_law_1.01,0.251366400718689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,2,power_law_1.01,0.25442559719085694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,2,power_law_1.01,0.25338239669799806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,2,power_law_1.01,0.25628159046173093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,2,power_law_1.01,0.26823039054870607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,2,power_law_1.01,0.2946432113647461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,2,power_law_1.01,0.32586240768432617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,2,power_law_1.01,0.3759680032730103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,2,power_law_1.01,0.0648639976978302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,2,power_law_1.01,0.05149440169334411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,2,power_law_1.01,0.435532808303833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,2,power_law_1.01,0.05807999968528747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,2,power_law_1.01,0.05120000243186951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,2,power_law_1.01,0.5391359806060791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,2,power_law_1.01,0.06947199702262878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,2,power_law_1.01,0.0639680027961731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,2,power_law_1.01,0.6288896083831788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,2,power_law_1.01,0.09185280203819275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,2,power_law_1.01,0.08414080142974853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,2,power_law_1.01,0.8475008010864258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,2,power_law_1.01,0.1167680025100708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,2,power_law_1.01,0.09694719910621644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,2,power_law_1.01,1.0525568008422852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,2,power_law_1.01,0.15699199438095093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,2,power_law_1.01,0.13826559782028197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,2,power_law_1.01,1.4121472358703613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,2,power_law_1.01,0.16571520566940307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,2,power_law_1.01,0.14162559509277345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,2,power_law_1.01,1.8520320892333983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,2,power_law_1.01,0.16369279623031616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,2,power_law_1.01,0.1467967987060547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,2,power_law_1.01,2.145055961608887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,2,power_law_1.01,0.16189440488815307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,2,power_law_1.01,0.14510719776153563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,2,power_law_1.01,0.16323200464248658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,2,power_law_1.01,0.14851200580596924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,2,power_law_1.01,3.4526592254638673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,2,power_law_1.01,0.17000319957733154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,2,power_law_1.01,0.15285120010375977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,2,power_law_1.01,0.16871039867401122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,2,power_law_1.01,6.899590301513672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,2,power_law_1.01,0.15652480125427246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,2,power_law_1.01,0.1747712016105652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,2,power_law_1.01,0.16112639904022216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,2,power_law_1.01,0.17327359914779664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,2,power_law_1.01,0.15830399990081787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,2,power_law_1.01,0.17870080471038818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,2,power_law_1.01,0.17188479900360107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,2,power_law_1.01,0.1866047978401184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,2,power_law_1.01,0.16925439834594727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,2,power_law_1.01,0.19685759544372558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,2,power_law_1.01,0.18807040452957152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,2,power_law_1.01,0.21611518859863282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,2,power_law_1.01,0.20948479175567628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.01,0.24153599739074708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,2,power_law_1.01,0.23228158950805664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.01,0.28998401165008547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,2,power_law_1.01,0.2799232006072998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.01,0.3331968069076538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,2,power_law_1.01,0.32724480628967284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.01,0.41171841621398925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,2,power_law_1.01,0.43439998626708987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.01,0.5086207866668702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,2,power_law_1.01,0.48836479187011717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.01,0.6967936038970948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,2,power_law_1.01,0.6890624046325684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.01,0.906278419494629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,2,power_law_1.01,0.870143985748291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.01,1.2856575965881347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,2,power_law_1.01,1.243398380279541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.01,1.6018560409545899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,2,power_law_1.01,1.5699263572692872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,2,power_law_1.01,1.9184703826904297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,2,power_law_1.01,1.9979520797729493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,2,power_law_1.01,3.0130367279052734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,2,power_law_1.01,3.097868728637695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,2,power_law_1.01,6.068857574462891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,2,power_law_1.01,6.018745422363281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,2,power_law_1.01,0.021836799383163453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,2,power_law_1.01,0.024089600145816802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,2,power_law_1.01,0.03294720053672791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,2,power_law_1.01,0.038867199420928956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,2,power_law_1.01,0.046214398741722104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,2,power_law_1.01,0.05927039980888367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,2,power_law_1.01,0.06561920046806335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,2,power_law_1.01,0.06866559982299805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,2,power_law_1.01,0.06618239879608154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,2,power_law_1.01,0.06988160014152527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,2,power_law_1.01,0.07139840126037597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,2,power_law_1.01,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,2,power_law_1.01,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,2,power_law_1.01,0.07793279886245727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,2,power_law_1.01,0.08737279772758484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,2,power_law_1.01,0.08607360124588012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,2,power_law_1.01,0.08612480163574218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,2,power_law_1.01,0.15738879442214965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.01,0.17906559705734254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.01,0.14835200309753419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.01,0.1522047996520996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.01,0.17960959672927856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.01,0.20111360549926757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.01,0.26188158988952637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.01,0.3247999906539917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.01,0.44588799476623536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.01,0.5503359794616699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,2,power_law_1.01,0.7223167896270752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,2,power_law_1.01,1.1842495918273925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,2,power_law_1.01,2.1313472747802735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,2,power_law_1.2,0.07078400254249573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,2,power_law_1.2,0.06355199813842774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,2,power_law_1.2,0.07448959946632386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,2,power_law_1.2,0.11351679563522339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,2,power_law_1.2,0.14094719886779786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,2,power_law_1.2,0.22957439422607423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,2,power_law_1.2,0.23885440826416016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,2,power_law_1.2,0.2447360038757324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,2,power_law_1.2,0.2452159881591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,2,power_law_1.2,0.24713599681854248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,2,power_law_1.2,0.24968960285186767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,2,power_law_1.2,0.2561919927597046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,2,power_law_1.2,0.2592832088470459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,2,power_law_1.2,0.2613503932952881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,2,power_law_1.2,0.2651007890701294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,2,power_law_1.2,0.27714560031890867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,2,power_law_1.2,0.28654720783233645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,2,power_law_1.2,0.3164416074752808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,2,power_law_1.2,0.33897600173950193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,2,power_law_1.2,0.3952960014343262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,2,power_law_1.2,0.4546175956726074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,2,power_law_1.2,0.051974397897720334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,2,power_law_1.2,0.06618880033493042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,2,power_law_1.2,0.589792013168335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,2,power_law_1.2,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,2,power_law_1.2,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,2,power_law_1.2,0.6531199932098388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,2,power_law_1.2,0.06154879927635193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,2,power_law_1.2,0.06791679859161377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,2,power_law_1.2,0.8921216011047364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,2,power_law_1.2,0.08238080143928528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,2,power_law_1.2,0.08947839736938476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,2,power_law_1.2,1.126527976989746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,2,power_law_1.2,0.09349759817123413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,2,power_law_1.2,0.10560640096664428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,2,power_law_1.2,1.482476806640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,2,power_law_1.2,0.13628159761428832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,2,power_law_1.2,0.15858559608459472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,2,power_law_1.2,2.0855680465698243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,2,power_law_1.2,0.14874240159988403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,2,power_law_1.2,0.15961600542068483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,2,power_law_1.2,0.15167360305786132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,2,power_law_1.2,0.16654080152511597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,2,power_law_1.2,2.5288000106811523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,2,power_law_1.2,0.15202560424804687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,2,power_law_1.2,0.16993919610977173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,2,power_law_1.2,3.875584030151367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,2,power_law_1.2,0.1473471999168396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,2,power_law_1.2,0.17100160121917723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,2,power_law_1.2,0.16300159692764282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,2,power_law_1.2,0.17359999418258668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,2,power_law_1.2,7.951084899902344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,2,power_law_1.2,0.15808000564575195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,2,power_law_1.2,0.17567360401153564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,2,power_law_1.2,0.15946240425109864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,2,power_law_1.2,0.1778432011604309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,2,power_law_1.2,0.1695680022239685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,2,power_law_1.2,0.1791551947593689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,2,power_law_1.2,0.17490559816360474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,2,power_law_1.2,0.18254079818725585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,2,power_law_1.2,0.1752575993537903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,2,power_law_1.2,0.19290239810943605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,2,power_law_1.2,0.19313280582427977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,2,power_law_1.2,0.19845759868621826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,2,power_law_1.2,0.20981760025024415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,2,power_law_1.2,0.22792320251464843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,2,power_law_1.2,0.23778560161590576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.2,0.24391040802001954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,2,power_law_1.2,0.2889024019241333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.2,0.2855871915817261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,2,power_law_1.2,0.3432512044906616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.2,0.3264064073562622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,2,power_law_1.2,0.43050241470336914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.2,0.43540477752685547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,2,power_law_1.2,0.5548287868499756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.2,0.5261568069458008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,2,power_law_1.2,0.7131392002105713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.2,0.7296063899993896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,2,power_law_1.2,0.9192255973815918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.2,0.889094352722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.2,1.2670720100402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,2,power_law_1.2,1.3376128196716308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.2,1.7127872467041017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,2,power_law_1.2,1.6798528671264648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,2,power_law_1.2,2.1649728775024415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,2,power_law_1.2,1.9629056930541993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,2,power_law_1.2,3.083987236022949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,2,power_law_1.2,3.341356658935547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,2,power_law_1.2,6.309823989868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,2,power_law_1.2,7.252601623535156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,2,power_law_1.2,0.02157440036535263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,2,power_law_1.2,0.025536000728607178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,2,power_law_1.2,0.03041279911994934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,2,power_law_1.2,0.03880960047245026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,2,power_law_1.2,0.04177919924259186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,2,power_law_1.2,0.061286401748657224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,2,power_law_1.2,0.06435840129852295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,2,power_law_1.2,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,2,power_law_1.2,0.06741120219230652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,2,power_law_1.2,0.06913279891014099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,2,power_law_1.2,0.07565439939498901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,2,power_law_1.2,0.07530879974365234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,2,power_law_1.2,0.07685120105743408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,2,power_law_1.2,0.07685120105743408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,2,power_law_1.2,0.09181439876556396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,2,power_law_1.2,0.08339840173721313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,2,power_law_1.2,0.08803840279579163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,2,power_law_1.2,0.16288000345230103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.2,0.18194559812545777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.2,0.14552320241928102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.2,0.16068480014801026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.2,0.18440959453582764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.2,0.21752960681915284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.2,0.27048959732055666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.2,0.34226560592651367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.2,0.45845761299133303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.2,0.6087615966796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,2,power_law_1.2,0.7333888053894043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,2,power_law_1.2,1.1428863525390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,2,power_law_1.2,2.2586624145507814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,1,balanced,0.05635199944178263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,1,balanced,0.07382399837176006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,1,balanced,0.10196800033251445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,1,balanced,0.1586079994837443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,1,balanced,0.26662933826446533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,1,balanced,0.25413866837819415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,1,balanced,0.2548000017801921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,1,balanced,0.24887466430664062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,1,balanced,0.2627840042114258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,1,balanced,0.25891733169555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,1,balanced,0.2531733314196269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,1,balanced,0.26203733682632446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,1,balanced,0.2635253270467122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,1,balanced,0.2632906635602315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,1,balanced,0.26772799094518024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,1,balanced,0.2741439938545227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,1,balanced,0.2845919926961263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,1,balanced,0.30696000655492145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,1,balanced,0.3253013292948405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,1,balanced,0.0668693333864212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,1,balanced,0.07976533472537994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,1,balanced,0.04829333225886027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,1,balanced,0.3708053429921468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,1,balanced,0.10596799850463867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,1,balanced,0.05717866619427999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,1,balanced,0.15181333820025125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,1,balanced,0.07947733501593272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,1,balanced,0.40704532464345294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,1,balanced,0.2312800089518229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,1,balanced,0.11167466640472412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,1,balanced,0.2099626660346985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,1,balanced,0.1853546698888143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,1,balanced,0.17693867286046347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,1,balanced,0.2126506765683492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,1,balanced,0.17359467347462973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,1,balanced,0.2142826716105143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,1,balanced,0.5569866498311361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,1,balanced,0.17624000708262125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,1,balanced,0.21644800901412964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,1,balanced,0.1783519983291626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,1,balanced,0.21296000480651855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,1,balanced,0.17443732420603433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,1,balanced,0.2182240088780721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,1,balanced,0.1760480006535848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,1,balanced,0.6386133432388306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,1,balanced,0.21715732415517172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,1,balanced,0.1818880041440328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,1,balanced,0.22738667329152426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,1,balanced,0.18634132544199625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,1,balanced,0.22126400470733643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,1,balanced,0.1857973337173462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,1,balanced,0.23388266563415527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,1,balanced,0.19614932934443155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,1,balanced,0.24006932973861694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,1,balanced,0.8809813658396403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,1,balanced,0.20125333468119302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,1,balanced,0.24928534030914307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,1,balanced,0.2137440045674642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,1,balanced,0.2762879927953084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,1,balanced,0.2419040004412333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,1,balanced,0.29713600873947144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,1,balanced,0.26392533381779987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,1,balanced,1.1229066848754883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,1,balanced,0.36116798718770343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,1,balanced,0.31802666187286377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,1,balanced,0.406826655069987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,1,balanced,0.363813320795695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,1,balanced,0.6362080176671346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,1,balanced,0.5136160055796305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,1,balanced,1.6437439918518066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,1,balanced,0.7415626843770345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,1,balanced,0.6150346597035726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,1,balanced,1.0505066712697346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,1,balanced,0.01977066695690155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,1,balanced,0.022650666534900665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,1,balanced,0.024874667326609295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,1,balanced,0.8783573309580485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,1,balanced,0.03266133368015289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,1,balanced,0.05120000243186951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,1,balanced,2.1318079630533853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,1,balanced,0.05340266724427541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,1,balanced,0.054655998945236206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,1,balanced,1.3337440490722656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,1,balanced,0.05729066828886668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,1,balanced,0.05955199897289276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,1,balanced,0.06113066772619883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,1,balanced,0.06493333478768666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,1,balanced,1.1365119616190593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,1,balanced,0.06517333288987477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,1,balanced,0.0666293352842331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,1,balanced,0.07155733307202657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,1,balanced,0.06909866631031036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,1,balanced,0.07229333122571309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,1,balanced,1.925381342569987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,1,balanced,0.07540800174077351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,1,balanced,2.6493226687113443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,1,balanced,0.08218133449554443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,1,balanced,0.08950933814048767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,1,balanced,1.6414079666137695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,1,balanced,0.15200533469518027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,1,balanced,0.1750239928563436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,1,balanced,2.527381261189779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,1,balanced,0.16528532902399698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,1,balanced,0.17418134212493896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,1,balanced,2.1335946718851724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,1,balanced,0.2818133234977722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,1,balanced,4.144202550252278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,1,balanced,3.132842699686686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,1,balanced,0.29284799098968506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,1,balanced,2.645397345225016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,1,balanced,0.415450652440389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,1,balanced,0.5442293485005697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,1,balanced,4.875823974609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,1,balanced,4.157690684000651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,1,balanced,0.6708959738413492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,1,balanced,8.136330922444662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,1,balanced,1.069210688273112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,1,balanced,9.605978647867838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,1,balanced,8.188650767008463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,1,balanced,2.092917283376058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,1,power_law_1.01,0.0532800018787384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,1,power_law_1.01,0.06307200193405152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,1,power_law_1.01,0.08410239815711976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,1,power_law_1.01,0.11909120082855225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,1,power_law_1.01,0.1447808027267456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,1,power_law_1.01,0.21727359294891357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,1,power_law_1.01,0.22909440994262695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,1,power_law_1.01,0.2302016019821167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,1,power_law_1.01,0.2368704080581665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,1,power_law_1.01,0.24019200801849366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,1,power_law_1.01,0.24430720806121825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,1,power_law_1.01,0.24357120990753173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,1,power_law_1.01,0.2520256042480469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,1,power_law_1.01,0.259769606590271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,1,power_law_1.01,0.2612672090530396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,1,power_law_1.01,0.27307519912719724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,1,power_law_1.01,0.28974080085754395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,1,power_law_1.01,0.31715199947357176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,1,power_law_1.01,0.34163200855255127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,1,power_law_1.01,0.40776958465576174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,1,power_law_1.01,0.4680960178375244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,1,power_law_1.01,0.048256000876426695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,1,power_law_1.01,0.6025087833404541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,1,power_law_1.01,0.05252479910850525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,1,power_law_1.01,0.7173567771911621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,1,power_law_1.01,0.06673280000686646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,1,power_law_1.01,0.9643839836120606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,1,power_law_1.01,0.08737279772758484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,1,power_law_1.01,1.1996031761169434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,1,power_law_1.01,0.10607359409332276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,1,power_law_1.01,0.14784640073776245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,1,power_law_1.01,1.6898176193237304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,1,power_law_1.01,0.15299839973449708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,1,power_law_1.01,2.1811712265014647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,1,power_law_1.01,0.15635839700698853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,1,power_law_1.01,2.6904191970825195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,1,power_law_1.01,0.16097279787063598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,1,power_law_1.01,0.16264959573745727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,1,power_law_1.01,4.173542404174805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,1,power_law_1.01,0.16800639629364014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,1,power_law_1.01,8.139206695556641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,1,power_law_1.01,0.17424000501632692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,1,power_law_1.01,0.17828480005264283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,1,power_law_1.01,0.1786687970161438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,1,power_law_1.01,0.1891584038734436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,1,power_law_1.01,0.19906560182571412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,1,power_law_1.01,0.21329920291900634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,1,power_law_1.01,0.24759678840637206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,1,power_law_1.01,0.06060799956321716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,1,power_law_1.01,0.27834880352020264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,1,power_law_1.01,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,1,power_law_1.01,0.34024319648742674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,1,power_law_1.01,0.08788480162620545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,1,power_law_1.01,0.3976959943771362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,1,power_law_1.01,0.01847040057182312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,1,power_law_1.01,0.121561598777771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,1,power_law_1.01,0.5382783889770508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,1,power_law_1.01,0.021081599593162536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,1,power_law_1.01,0.13671040534973145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,1,power_law_1.01,0.6591040134429932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,1,power_law_1.01,0.025203201174736022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,1,power_law_1.01,0.195904004573822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,1,power_law_1.01,0.9150912284851074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,1,power_law_1.01,0.033606401085853575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,1,power_law_1.01,0.19633920192718507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,1,power_law_1.01,0.04639999866485596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,1,power_law_1.01,1.161081600189209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,1,power_law_1.01,0.20142080783843994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,1,power_law_1.01,0.05804160237312317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,1,power_law_1.01,0.20674560070037842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,1,power_law_1.01,1.6635648727416992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,1,power_law_1.01,0.06152960062026978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,1,power_law_1.01,0.20989439487457276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,1,power_law_1.01,2.157222366333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,1,power_law_1.01,0.06432639956474304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,1,power_law_1.01,0.20862081050872802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,1,power_law_1.01,2.6630016326904298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,1,power_law_1.01,0.06639360189437866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,1,power_law_1.01,0.21478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,1,power_law_1.01,0.06820480227470398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,1,power_law_1.01,4.150668716430664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,1,power_law_1.01,0.21758720874786378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,1,power_law_1.01,0.07537919878959656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,1,power_law_1.01,0.22045440673828126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,1,power_law_1.01,8.182860565185546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,1,power_law_1.01,0.07414399981498718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,1,power_law_1.01,0.2320255994796753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,1,power_law_1.01,0.07613440155982971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,1,power_law_1.01,0.23600640296936035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,1,power_law_1.01,0.08235520124435425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,1,power_law_1.01,0.2486975908279419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,1,power_law_1.01,0.0794048011302948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,1,power_law_1.01,0.2980288028717041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,1,power_law_1.01,0.08118399977684021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.01,0.3331199884414673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,1,power_law_1.01,0.08971520066261292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.01,0.41351680755615233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,1,power_law_1.01,0.09525759816169739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.01,0.4824319839477539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.01,0.1031999945640564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.01,0.6390336036682129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.01,0.1962496042251587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.01,0.7858496189117432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.01,0.22753920555114746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.01,1.0841024398803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.01,0.18921600580215453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.01,1.3748031616210938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.01,0.2167743921279907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.01,1.9633024215698243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.01,0.27716479301452634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.01,0.33930239677429197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.01,2.528678321838379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.01,0.4621823787689209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.01,3.1183744430541993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.01,0.5860735893249511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,1,power_law_1.01,4.853657531738281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.01,0.7094848155975342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,1,power_law_1.01,1.0863231658935546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,1,power_law_1.01,9.525843048095703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,1,power_law_1.01,2.1108800888061525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,1,power_law_1.2,0.054425597190856934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,1,power_law_1.2,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,1,power_law_1.2,0.07486720085144043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,1,power_law_1.2,0.10963200330734253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,1,power_law_1.2,0.12447359561920165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,1,power_law_1.2,0.2153856039047241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,1,power_law_1.2,0.23348479270935057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,1,power_law_1.2,0.23971199989318848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,1,power_law_1.2,0.2354367971420288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,1,power_law_1.2,0.24661118984222413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,1,power_law_1.2,0.24927999973297119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,1,power_law_1.2,0.2521791934967041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,1,power_law_1.2,0.2554624080657959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,1,power_law_1.2,0.2614527940750122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,1,power_law_1.2,0.26803200244903563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,1,power_law_1.2,0.27942399978637694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,1,power_law_1.2,0.28988161087036135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,1,power_law_1.2,0.321612811088562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,1,power_law_1.2,0.35387520790100097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,1,power_law_1.2,0.41173758506774905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,1,power_law_1.2,0.4691328048706055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,1,power_law_1.2,0.6013120174407959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,1,power_law_1.2,0.7173567771911621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,1,power_law_1.2,0.9677824020385742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,1,power_law_1.2,1.2006143569946288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,1,power_law_1.2,1.6835519790649414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,1,power_law_1.2,2.166828727722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,1,power_law_1.2,2.680723190307617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,1,power_law_1.2,4.162956619262696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,1,power_law_1.2,0.048256000876426695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,1,power_law_1.2,0.051641601324081424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,1,power_law_1.2,8.164595031738282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,1,power_law_1.2,0.0636352002620697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,1,power_law_1.2,0.08664320111274719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,1,power_law_1.2,0.09582080245018006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,1,power_law_1.2,0.15244799852371216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,1,power_law_1.2,0.15538560152053832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,1,power_law_1.2,0.16303999423980714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,1,power_law_1.2,0.16391680240631104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,1,power_law_1.2,0.1661120057106018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,1,power_law_1.2,0.16908799409866332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,1,power_law_1.2,0.17270400524139404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,1,power_law_1.2,0.17935359477996826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,1,power_law_1.2,0.019148799777030944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,1,power_law_1.2,0.18364800214767457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,1,power_law_1.2,0.02077440023422241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,1,power_law_1.2,0.19827200174331666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,1,power_law_1.2,0.024460799992084503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,1,power_law_1.2,0.2036992073059082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,1,power_law_1.2,0.0308351993560791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,1,power_law_1.2,0.21615359783172608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,1,power_law_1.2,0.041305598616600034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,1,power_law_1.2,0.25138559341430666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,1,power_law_1.2,0.06538239717483521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,1,power_law_1.2,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,1,power_law_1.2,0.27882239818572996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,1,power_law_1.2,0.06680319905281067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,1,power_law_1.2,0.06183040142059326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,1,power_law_1.2,0.3448319911956787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,1,power_law_1.2,0.06332160234451294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,1,power_law_1.2,0.08814079761505127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,1,power_law_1.2,0.40133118629455566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,1,power_law_1.2,0.06762880086898804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,1,power_law_1.2,0.10512000322341919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,1,power_law_1.2,0.5372032165527344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,1,power_law_1.2,0.07062399983406067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,1,power_law_1.2,0.12954879999160768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,1,power_law_1.2,0.6576704025268555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,1,power_law_1.2,0.07383679747581481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,1,power_law_1.2,0.18654719591140748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,1,power_law_1.2,0.9157183647155762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,1,power_law_1.2,0.07427840232849121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,1,power_law_1.2,0.20090880393981933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,1,power_law_1.2,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,1,power_law_1.2,1.1593152046203614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,1,power_law_1.2,0.20830719470977782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,1,power_law_1.2,0.0809216022491455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,1,power_law_1.2,0.20578560829162598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,1,power_law_1.2,1.6498367309570312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,1,power_law_1.2,0.08094080090522766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,1,power_law_1.2,0.21092479228973388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,1,power_law_1.2,2.150668716430664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,1,power_law_1.2,0.08144000172615051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,1,power_law_1.2,0.21790719032287598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,1,power_law_1.2,2.652012825012207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,1,power_law_1.2,0.08917760252952575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,1,power_law_1.2,0.21340160369873046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,1,power_law_1.2,0.0949567973613739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,1,power_law_1.2,0.2207551956176758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,1,power_law_1.2,4.141849517822266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.2,0.10143359899520873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,1,power_law_1.2,0.2213887929916382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,1,power_law_1.2,8.145497894287109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.2,0.19576319456100463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,1,power_law_1.2,0.23543040752410888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.2,0.2268160104751587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,1,power_law_1.2,0.24481918811798095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.2,0.1934399962425232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,1,power_law_1.2,0.2598016023635864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.2,0.22430078983306884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,1,power_law_1.2,0.3024512052536011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.2,0.3339967966079712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.2,0.2830336093902588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.2,0.4142399787902832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.2,0.3424639940261841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.2,0.48297600746154784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.2,0.4638656139373779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.2,0.6443327903747559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.2,0.5848320007324219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.2,0.7858496189117432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.2,0.71495680809021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.2,1.0781824111938476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,1,power_law_1.2,1.0937600135803223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.2,1.3645183563232421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,1,power_law_1.2,2.10949764251709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.2,1.94400634765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.2,2.5208831787109376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.2,3.1031871795654298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,1,power_law_1.2,4.834495925903321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,1,power_law_1.2,9.48018569946289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,4,balanced,0.09537600477536519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,4,balanced,0.1402133305867513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,4,balanced,0.21090133984883627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,4,balanced,0.33581332365671795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,4,balanced,0.5926293134689331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,4,balanced,0.746837298075358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,4,balanced,0.7145013014475504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,4,balanced,0.7269759972890218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,4,balanced,0.7831093470255533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,4,balanced,0.7367146809895834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,4,balanced,0.7206079959869385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,4,balanced,0.7179253101348877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,4,balanced,0.7378186384836832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,4,balanced,0.7152586778004965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,4,balanced,0.7404692967732748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,4,balanced,0.7304960091908773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,4,balanced,0.749946673711141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,4,balanced,0.7173013687133789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,4,balanced,0.7083040078481039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,4,balanced,0.7116639614105225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,4,balanced,0.7199199994405111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,4,balanced,1.271008014678955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,4,balanced,1.1326346397399902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,4,balanced,1.6965120633443196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,4,balanced,0.07150400181611379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,4,balanced,0.09822400410970052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,4,balanced,0.13105600078900656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,4,balanced,0.20068265994389853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,4,balanced,0.3280373414357503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,4,balanced,0.4039786656697591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,4,balanced,0.40294400850931805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,4,balanced,2.272736072540283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,4,balanced,0.39481067657470703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,4,balanced,0.41257067521413165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,4,balanced,0.4052160183588664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,4,balanced,0.4099946816762288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,4,balanced,0.42365864912668866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,4,balanced,0.4175413449605306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,4,balanced,0.4234773317972819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,4,balanced,0.41174399852752686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,4,balanced,0.07457066575686137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,4,balanced,2.3743359247843423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,4,balanced,0.4272213379542033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,4,balanced,0.1104906698067983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,4,balanced,0.41655464967091876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,4,balanced,0.13731200496355692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,4,balanced,0.2446933388710022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,4,balanced,0.4203679958979289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,4,balanced,0.37587201595306396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,4,balanced,0.3880266745885213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,4,balanced,0.4319146474202474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,4,balanced,0.3927520116170247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,4,balanced,0.37089598178863525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,4,balanced,0.43486400445302326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,4,balanced,0.3645013173421224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,4,balanced,0.04483200112978617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,4,balanced,0.38599467277526855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,4,balanced,0.05282133320967356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,4,balanced,0.4599573214848836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,4,balanced,0.3856319983800252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,4,balanced,0.06922666728496552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,4,balanced,3.510165214538574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,4,balanced,0.37833066781361896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,4,balanced,0.09462933739026387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,4,balanced,0.3924853404362996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,4,balanced,0.14525866508483887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,4,balanced,0.7663040161132812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,4,balanced,0.17831466595331827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,4,balanced,0.38767464955647785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,4,balanced,0.17880533138910928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,4,balanced,0.3890186548233032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,4,balanced,0.18343466520309448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,4,balanced,0.40355201562245685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,4,balanced,0.1878933310508728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,4,balanced,0.7198560237884521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,4,balanced,0.21406400203704834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,4,balanced,0.3967039982477824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,4,balanced,0.21818667650222778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,4,balanced,0.4115519920984904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,4,balanced,0.2190826733907064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,4,balanced,0.31107733647028607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,4,balanced,0.3946880102157593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,4,balanced,0.31487466891606647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,4,balanced,1.0859146912892659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,4,balanced,0.31042667229970294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,4,balanced,0.3844746748606364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,4,balanced,3.590730667114258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,4,balanced,0.3957333167394002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,4,balanced,0.43263999621073407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,4,balanced,0.39636266231536865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,4,balanced,0.4204639991124471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,4,balanced,1.412943998972575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,4,balanced,0.4357440074284871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,4,balanced,0.7287253538767496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,4,balanced,0.46621867020924884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,4,balanced,0.6945280234018961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,4,balanced,0.5018346707026163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,4,balanced,0.5493493477503458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,4,balanced,1.6064799626668294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,4,balanced,1.008133331934611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,4,balanced,0.5829759836196899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,4,balanced,5.7809492746988935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,4,balanced,0.9748586813608805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,4,balanced,1.3151626586914062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,4,balanced,2.3357226053873696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,4,balanced,1.0889920393625896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,4,balanced,1.4727840423583984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,4,balanced,1.5992693901062012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,4,balanced,2.449199994405111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,4,balanced,2.1559200286865234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,4,balanced,2.069215933481852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,4,balanced,3.9392693837483725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,4,balanced,11.685498555501303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,4,balanced,2.265183925628662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,4,balanced,2.2463040351867676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,4,balanced,3.614442825317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,4,balanced,3.856202761332194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,4,balanced,7.860250473022461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,4,balanced,7.13913091023763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,4,balanced,7.244501113891602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,4,power_law_1.01,0.21650559902191163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,4,power_law_1.01,0.2592704057693481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,4,power_law_1.01,0.18619519472122192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,4,power_law_1.01,0.3110399961471558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,4,power_law_1.01,0.36376960277557374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,4,power_law_1.01,0.5974271774291993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,4,power_law_1.01,0.6129856109619141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,4,power_law_1.01,0.6368512153625489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,4,power_law_1.01,0.6609344005584716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,4,power_law_1.01,0.6439487934112549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,4,power_law_1.01,0.66211838722229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,4,power_law_1.01,0.6995967864990235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,4,power_law_1.01,0.6815872192382812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,4,power_law_1.01,0.6941440105438232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,4,power_law_1.01,0.6852416038513184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,4,power_law_1.01,0.6961088180541992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,4,power_law_1.01,0.7552896022796631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,4,power_law_1.01,0.813811206817627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,4,power_law_1.01,0.8605695724487304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,4,power_law_1.01,0.926534366607666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,4,power_law_1.01,1.139411163330078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,4,power_law_1.01,1.364896011352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,4,power_law_1.01,1.5720895767211913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,4,power_law_1.01,1.941676712036133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,4,power_law_1.01,2.5119871139526366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,4,power_law_1.01,3.43554573059082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,4,power_law_1.01,4.469311904907227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,4,power_law_1.01,5.010271835327148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,4,power_law_1.01,7.796275329589844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,4,power_law_1.01,15.861503601074219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,4,power_law_1.01,0.1366719961166382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,4,power_law_1.01,0.17419519424438476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,4,power_law_1.01,0.12494080066680908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,4,power_law_1.01,0.211411190032959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,4,power_law_1.01,0.230515193939209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,4,power_law_1.01,0.3467456102371216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,4,power_law_1.01,0.35676159858703616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,4,power_law_1.01,0.35742080211639404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,4,power_law_1.01,0.38442239761352537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,4,power_law_1.01,0.3880000114440918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,4,power_law_1.01,0.387993597984314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,4,power_law_1.01,0.3787071943283081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,4,power_law_1.01,0.39032320976257323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,4,power_law_1.01,0.3975744009017944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,4,power_law_1.01,0.4231423854827881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,4,power_law_1.01,0.42739200592041016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,4,power_law_1.01,0.4320064067840576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,4,power_law_1.01,0.4720448017120361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,4,power_law_1.01,0.5078207969665527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,4,power_law_1.01,0.5790080070495606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,4,power_law_1.01,0.691974401473999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,4,power_law_1.01,0.829804801940918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,4,power_law_1.01,0.14447360038757323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,4,power_law_1.01,0.9552127838134765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,4,power_law_1.01,0.1856384038925171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,4,power_law_1.01,1.2075136184692383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,4,power_law_1.01,0.11720319986343383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,4,power_law_1.01,1.5421952247619628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,4,power_law_1.01,0.18391040563583375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,4,power_law_1.01,2.0525312423706055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,4,power_law_1.01,0.23505918979644774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,4,power_law_1.01,0.050374400615692136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,4,power_law_1.01,0.3407167911529541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,4,power_law_1.01,2.6836095809936524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,4,power_law_1.01,0.06239359974861145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,4,power_law_1.01,0.35540480613708497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,4,power_law_1.01,0.05434880256652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,4,power_law_1.01,3.5217216491699217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,4,power_law_1.01,0.3709887981414795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,4,power_law_1.01,0.08372480273246766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,4,power_law_1.01,0.3843391895294189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,4,power_law_1.01,5.41338882446289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,4,power_law_1.01,0.10431360006332398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,4,power_law_1.01,0.3689215898513794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,4,power_law_1.01,0.1428096055984497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,4,power_law_1.01,9.725888061523438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,4,power_law_1.01,0.37817599773406985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,4,power_law_1.01,0.14933120012283324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,4,power_law_1.01,0.3690560102462769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,4,power_law_1.01,0.1553984045982361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,4,power_law_1.01,0.3720704078674316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,4,power_law_1.01,0.1542464017868042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,4,power_law_1.01,0.3830912113189697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,4,power_law_1.01,0.16903040409088135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,4,power_law_1.01,0.3993535995483398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,4,power_law_1.01,0.16728960275650023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,4,power_law_1.01,0.3975104093551636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,4,power_law_1.01,0.16947840452194213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,4,power_law_1.01,0.41866240501403806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,4,power_law_1.01,0.2119296073913574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,4,power_law_1.01,0.4469312191009521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,4,power_law_1.01,0.19435520172119142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.01,0.5148287773132324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,4,power_law_1.01,0.1944640040397644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.01,0.5548863887786866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,4,power_law_1.01,0.340499210357666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,4,power_law_1.01,0.3696000099182129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.01,0.632531213760376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,4,power_law_1.01,0.3188159942626953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.01,0.7472511768341065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.01,0.32128000259399414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.01,0.9112064361572265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.01,0.34841599464416506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.01,1.2100095748901367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.01,0.35018880367279054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.01,1.5134783744812013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.01,0.41765761375427246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.01,1.9913600921630858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.01,0.4802879810333252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.01,2.6674623489379883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.01,0.6113984107971191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,4,power_law_1.01,3.146419143676758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.01,0.7880640029907227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,4,power_law_1.01,5.201523208618164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.01,1.1346367835998534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.01,1.393939208984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,4,power_law_1.01,9.223353576660156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,4,power_law_1.01,1.6667200088500977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,4,power_law_1.01,2.5371776580810548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,4,power_law_1.01,5.003007888793945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,4,power_law_1.2,0.21597440242767335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,4,power_law_1.2,0.1659775972366333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,4,power_law_1.2,0.19413119554519653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,4,power_law_1.2,0.2613823890686035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,4,power_law_1.2,0.3475327968597412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,4,power_law_1.2,0.5800640106201171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,4,power_law_1.2,0.6107007980346679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,4,power_law_1.2,0.6225728034973145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,4,power_law_1.2,0.6562751770019531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,4,power_law_1.2,0.6676159858703613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,4,power_law_1.2,0.6753471851348877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,4,power_law_1.2,0.6940480232238769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,4,power_law_1.2,0.7178624153137207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,4,power_law_1.2,0.7080704212188721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,4,power_law_1.2,0.7352255821228028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,4,power_law_1.2,0.768179178237915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,4,power_law_1.2,0.7900608062744141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,4,power_law_1.2,0.8193216323852539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,4,power_law_1.2,0.8781503677368164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,4,power_law_1.2,1.0359487533569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,4,power_law_1.2,1.1407615661621093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,4,power_law_1.2,1.4574463844299317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,4,power_law_1.2,1.7914815902709962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,4,power_law_1.2,2.213235282897949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,4,power_law_1.2,2.5652288436889648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,4,power_law_1.2,3.787968063354492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,4,power_law_1.2,5.236435317993164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,4,power_law_1.2,6.431980895996094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,4,power_law_1.2,9.519667053222657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,4,power_law_1.2,18.697836303710936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,4,power_law_1.2,0.13452160358428955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,4,power_law_1.2,0.0986624002456665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,4,power_law_1.2,0.13507200479507447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,4,power_law_1.2,0.16812800168991088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,4,power_law_1.2,0.2128767967224121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,4,power_law_1.2,0.3353856086730957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,4,power_law_1.2,0.35819520950317385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,4,power_law_1.2,0.379423999786377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,4,power_law_1.2,0.3864448070526123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,4,power_law_1.2,0.37922561168670654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,4,power_law_1.2,0.3849407911300659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,4,power_law_1.2,0.4036863803863525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,4,power_law_1.2,0.38022398948669434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,4,power_law_1.2,0.4165952205657959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,4,power_law_1.2,0.41032958030700684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,4,power_law_1.2,0.4350272178649902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,4,power_law_1.2,0.4473599910736084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,4,power_law_1.2,0.15515520572662353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,4,power_law_1.2,0.49502081871032716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,4,power_law_1.2,0.09955840110778809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,4,power_law_1.2,0.554259204864502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,4,power_law_1.2,0.12377599477767945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,4,power_law_1.2,0.6180287837982178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,4,power_law_1.2,0.17497600317001344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,4,power_law_1.2,0.7055103778839111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,4,power_law_1.2,0.2133568048477173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,4,power_law_1.2,0.8510592460632325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,4,power_law_1.2,0.33087360858917236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,4,power_law_1.2,0.3480767965316772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,4,power_law_1.2,1.049619197845459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,4,power_law_1.2,0.3743168115615845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,4,power_law_1.2,1.3431296348571777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,4,power_law_1.2,0.3666176080703735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,4,power_law_1.2,1.6000192642211915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,4,power_law_1.2,0.05073919892311096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,4,power_law_1.2,0.38035199642181394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,4,power_law_1.2,2.476608085632324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,4,power_law_1.2,0.04624640047550201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,4,power_law_1.2,0.38255360126495364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,4,power_law_1.2,3.0439935684204102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,4,power_law_1.2,0.0587007999420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,4,power_law_1.2,0.386246395111084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,4,power_law_1.2,0.06908159852027893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,4,power_law_1.2,0.3976448059082031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,4,power_law_1.2,0.08922240138053894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,4,power_law_1.2,0.4117119789123535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,4,power_law_1.2,0.1397760033607483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,4,power_law_1.2,0.42050561904907224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,4,power_law_1.2,0.14910080432891845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,4,power_law_1.2,0.4299647808074951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,4,power_law_1.2,0.15210239887237548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,4,power_law_1.2,0.4309247970581055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,4,power_law_1.2,0.15953919887542725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,4,power_law_1.2,0.4759103775024414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,4,power_law_1.2,0.16572799682617187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.2,0.5166592121124267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,4,power_law_1.2,0.16429439783096314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.2,0.584876823425293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,4,power_law_1.2,0.1700160026550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.2,0.6662015914916992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,4,power_law_1.2,0.2038719892501831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.2,0.8288703918457031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,4,power_law_1.2,0.21594879627227784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.2,0.9935296058654786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,4,power_law_1.2,0.19839359521865846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.2,1.2809791564941406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,4,power_law_1.2,0.3418175935745239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.2,1.6925119400024413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,4,power_law_1.2,0.3845760107040405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.2,2.237945556640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,4,power_law_1.2,0.3263744115829468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.2,3.0118656158447266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.2,0.3296447992324829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,4,power_law_1.2,3.8826942443847656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.2,0.3452159881591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.2,0.36830079555511475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,4,power_law_1.2,5.035033416748047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.2,0.4502399921417236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.2,0.4995840072631836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,4,power_law_1.2,10.428160095214844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.2,0.6585792064666748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.2,0.8637503623962403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.2,1.1410880088806152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.2,1.4658368110656739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,4,power_law_1.2,1.8922943115234374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,4,power_law_1.2,2.820524787902832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,4,power_law_1.2,5.51141128540039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,2,balanced,0.09790933132171631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,2,balanced,0.1393173336982727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,2,balanced,0.21914132436116537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,2,balanced,0.3272586663564046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,2,balanced,0.5878026485443115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,2,balanced,0.7055359681447347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,2,balanced,0.6962080001831055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,2,balanced,0.6979733308156332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,2,balanced,0.7142773469289144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,2,balanced,0.7210986614227295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,2,balanced,0.6892639795939127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,2,balanced,0.7329920132954916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,2,balanced,0.7206079959869385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,2,balanced,0.6924853324890137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,2,balanced,0.7193280061086019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,2,balanced,0.715557336807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,2,balanced,0.7017172972361246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,2,balanced,0.7239147027333578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,2,balanced,0.720682700475057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,2,balanced,0.7353173096974691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,2,balanced,0.7649973233540853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,2,balanced,1.258239984512329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,2,balanced,1.1761226654052734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,2,balanced,1.7663733164469402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,2,balanced,2.301695982615153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,2,balanced,2.6385653813680015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,2,balanced,0.07282133400440216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,4,power_law_1.2,3.9299583435058594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,2,balanced,0.09582400321960449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,2,balanced,0.1304746667544047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,2,balanced,0.20090667406717935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,4,power_law_1.2,5.789056015014649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,2,balanced,0.33505598704020184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,2,balanced,0.3866879940032959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,4,power_law_1.2,11.0129150390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,2,balanced,0.3832106590270996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,2,balanced,0.37795201937357586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,2,balanced,0.38735465208689374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,2,balanced,3.784261385599772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,2,balanced,0.38203732172648114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,2,balanced,0.38440533479054767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,2,balanced,0.38733331362406415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,2,balanced,0.38228265444437665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,2,balanced,0.4221973419189453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,2,balanced,0.38924264907836914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,2,balanced,0.4028213421503703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,2,balanced,0.4020800193150838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,2,balanced,0.4296693404515584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,2,balanced,0.08122666676839192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,2,balanced,0.1006773312886556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,2,balanced,0.43828801314036053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,2,balanced,0.14179733395576477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,2,balanced,4.0218079884847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,2,balanced,0.22873065869013467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,2,balanced,0.35501333077748615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,2,balanced,0.47781864802042645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,2,balanced,0.41238399346669513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,2,balanced,0.38510934511820477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,2,balanced,0.5153973499933878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,2,balanced,0.3924320141474406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,2,balanced,0.38674132029215497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,2,balanced,0.377461314201355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,2,balanced,0.3860960006713867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,2,balanced,0.8033226331075033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,2,balanced,0.4127360184987386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,2,balanced,0.38739200433095294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,2,balanced,0.4108106692632039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,2,balanced,0.41074132919311523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,2,balanced,0.8066826661427816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,2,balanced,0.39848001797993976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,2,balanced,0.426474650700887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,2,balanced,0.41331199804941815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,2,balanced,1.1927200158437092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,2,balanced,0.4098506768544515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,2,balanced,0.436245322227478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,2,balanced,6.60865592956543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,2,balanced,0.46348798274993896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,2,balanced,1.5561866760253906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,2,balanced,0.7881653308868408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,2,balanced,0.8159093062082926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,2,balanced,1.9332000414530437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,2,balanced,1.147605339686076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,2,balanced,1.5230132738749187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,2,balanced,2.6517440478006997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,2,balanced,1.8432265917460124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,2,balanced,13.327765146891275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,2,balanced,3.0163679122924805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,2,balanced,2.575530687967936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,2,balanced,2.849498748779297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,2,balanced,0.03266666581233343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,2,balanced,0.04450666904449463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,2,balanced,0.06285333136717479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,2,balanced,0.08927999933560689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,2,balanced,4.821690559387207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,2,balanced,0.1434879998366038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,2,balanced,0.17628266414006552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,2,balanced,0.1824586590131124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,2,balanced,0.18956265846888223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,2,balanced,0.19716266791025797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,2,balanced,0.18766933679580688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,2,balanced,0.19403733809789023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,2,balanced,0.20339733362197876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,2,balanced,0.21287467082341513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,2,balanced,0.21535466114679971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,2,balanced,0.21779733896255493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,2,balanced,0.301530659198761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,2,balanced,4.563808123270671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,2,balanced,0.29283734162648517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,2,balanced,0.4437119960784912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,2,balanced,0.4777973492940267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,2,balanced,0.4396053155263265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,2,balanced,0.453546682993571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,2,balanced,0.47657068570454914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,2,balanced,0.47626666227976483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,2,balanced,0.8090346654256185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,2,balanced,0.8567093213399252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,2,balanced,9.55185063680013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,2,balanced,1.2500533262888591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,2,balanced,9.033781051635742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,2,balanced,1.6251947085062664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,2,power_law_1.01,0.1431615948677063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,2,power_law_1.01,0.2007040023803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,2,power_law_1.01,0.17604479789733887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,2,balanced,1.6991893450419109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,2,power_law_1.01,0.2683903932571411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,2,power_law_1.01,0.3269695997238159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,2,power_law_1.01,0.5444096088409424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,2,power_law_1.01,0.5778751850128174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,2,power_law_1.01,0.642406415939331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,2,power_law_1.01,0.601804780960083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,2,power_law_1.01,0.6296127796173095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,2,power_law_1.01,0.6524288177490234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,2,power_law_1.01,0.6691648006439209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,2,power_law_1.01,0.6600575923919678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,2,power_law_1.01,0.6699135780334473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,2,power_law_1.01,0.6777984142303467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,2,power_law_1.01,0.6877120018005372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,2,power_law_1.01,0.7109695911407471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,2,power_law_1.01,0.752678394317627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,2,power_law_1.01,0.7877888202667236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,2,power_law_1.01,0.9182463645935058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,2,power_law_1.01,1.0623552322387695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,2,balanced,2.916725476582845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,2,power_law_1.01,1.3045951843261718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,2,power_law_1.01,1.47957763671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,2,power_law_1.01,1.8880128860473633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,2,power_law_1.01,2.4214271545410155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,2,power_law_1.01,3.224319839477539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,2,power_law_1.01,3.9584766387939454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,2,power_law_1.01,5.272294235229492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,2,power_law_1.01,7.985919952392578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,2,power_law_1.01,14.94241943359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,2,balanced,5.361429214477539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,2,power_law_1.01,0.09664000272750854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,2,power_law_1.01,0.12518399953842163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,2,power_law_1.01,0.11449600458145141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,2,power_law_1.01,0.16609280109405516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,2,power_law_1.01,0.20721280574798584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,2,power_law_1.01,0.3304703950881958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,2,power_law_1.01,0.33031039237976073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,2,power_law_1.01,0.3376960039138794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,2,power_law_1.01,0.36110079288482666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,2,power_law_1.01,0.35671679973602294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,2,power_law_1.01,0.35827839374542236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,2,power_law_1.01,0.37150719165802004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,2,power_law_1.01,0.37881600856781006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,2,power_law_1.01,0.10857599973678589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,2,power_law_1.01,0.38487679958343507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,2,power_law_1.01,0.12255359888076782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,2,power_law_1.01,0.3890944004058838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,2,power_law_1.01,0.11552640199661254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,2,power_law_1.01,0.3862720012664795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,2,power_law_1.01,0.17526400089263916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,2,power_law_1.01,0.4205376148223877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,2,power_law_1.01,0.23405439853668214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,2,power_law_1.01,0.4469632148742676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,2,power_law_1.01,0.34469759464263916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,2,power_law_1.01,0.49965438842773435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,2,power_law_1.01,0.3635967969894409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,2,power_law_1.01,0.5702271938323975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,2,power_law_1.01,0.35996160507202146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,2,power_law_1.01,0.668665599822998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,2,power_law_1.01,0.36681599617004396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,2,power_law_1.01,0.8101632118225097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,2,power_law_1.01,0.38161280155181887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,2,power_law_1.01,0.9907072067260743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,2,power_law_1.01,0.3886656045913696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,2,power_law_1.01,1.3059391975402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,2,power_law_1.01,0.38064000606536863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,2,power_law_1.01,1.6166208267211915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,2,power_law_1.01,0.3877376079559326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,2,power_law_1.01,2.1242752075195312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,2,power_law_1.01,0.39971199035644533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,2,power_law_1.01,2.8631872177124023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,2,power_law_1.01,0.4027711868286133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,2,power_law_1.01,3.3679744720458986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,2,power_law_1.01,0.41283202171325684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,2,power_law_1.01,0.4167488098144531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,2,power_law_1.01,5.2557121276855465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,2,power_law_1.01,0.4506944179534912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.01,0.5237631797790527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,2,power_law_1.01,10.62520980834961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.01,0.5796351909637452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.01,0.6355455875396728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.01,0.7988992214202881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.01,0.9646976470947266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.01,1.2563839912414552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.01,1.5642239570617675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.01,2.1601343154907227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.01,2.717919921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,2,power_law_1.01,3.288288116455078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,2,power_law_1.01,5.277683258056641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,2,power_law_1.01,9.995116424560546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,2,power_law_1.01,0.0313728004693985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,2,power_law_1.01,0.04901120066642761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,2,power_law_1.01,0.04867199957370758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,2,power_law_1.01,0.07252479791641235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,2,power_law_1.01,0.08797439932823181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,2,power_law_1.01,0.13741439580917358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,2,power_law_1.01,0.14675840139389038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,2,power_law_1.01,0.14814079999923707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,2,power_law_1.01,0.15392639636993408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,2,power_law_1.01,0.15754879713058473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,2,power_law_1.01,0.1609984040260315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,2,power_law_1.01,0.16388479471206666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,2,power_law_1.01,0.16656639575958251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,2,power_law_1.01,0.16954879760742186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,2,power_law_1.01,0.18324480056762696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,2,power_law_1.01,0.19634560346603394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,2,power_law_1.01,0.19827200174331666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,2,power_law_1.01,0.38235518932342527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.01,0.4226111888885498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.01,0.3187455892562866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.01,0.3335423946380615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.01,0.37996160984039307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.01,0.42538881301879883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.01,0.545139217376709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.01,0.6665279865264893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.01,0.9210176467895508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,2,power_law_1.2,0.1460927963256836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.01,1.158956813812256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,2,power_law_1.2,0.11843199729919433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,2,power_law_1.01,1.4420096397399902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,2,power_law_1.2,0.16459519863128663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,2,power_law_1.01,2.070022392272949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,2,power_law_1.2,0.2362816095352173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,2,power_law_1.2,0.3247616052627563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,2,power_law_1.01,4.034150314331055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,2,power_law_1.2,0.5630335807800293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,2,power_law_1.2,0.5920127868652344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,2,power_law_1.2,0.6211328029632568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,2,power_law_1.2,0.6262464046478271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,2,power_law_1.2,0.674393606185913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,2,power_law_1.2,0.6649216175079345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,2,power_law_1.2,0.6791935920715332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,2,power_law_1.2,0.6987711906433105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,2,power_law_1.2,0.7082240104675293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,2,power_law_1.2,0.7019711971282959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,2,power_law_1.2,0.7094207763671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,2,power_law_1.2,0.7668223857879639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,2,power_law_1.2,0.8195136070251465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,2,power_law_1.2,0.863929557800293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,2,power_law_1.2,0.9732288360595703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,2,power_law_1.2,1.0893695831298829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,2,power_law_1.2,1.3735679626464843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,2,power_law_1.2,1.522163200378418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,2,power_law_1.2,1.978963279724121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,2,power_law_1.2,2.4847232818603517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,2,power_law_1.2,3.5664447784423827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,2,power_law_1.2,4.230169677734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,2,power_law_1.2,0.09525120258331299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,2,power_law_1.2,5.344889450073242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,2,power_law_1.2,0.09037439823150635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,2,power_law_1.2,8.080198669433594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,2,power_law_1.2,0.09792640209197997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,2,power_law_1.2,0.14712320566177367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,2,power_law_1.2,0.184607994556427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,2,power_law_1.2,16.463410949707033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,2,power_law_1.2,0.3186367988586426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,2,power_law_1.2,0.3304703950881958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,2,power_law_1.2,0.3512320041656494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,2,power_law_1.2,0.3696255922317505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,2,power_law_1.2,0.37011840343475344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,2,power_law_1.2,0.3793728113174438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,2,power_law_1.2,0.38380799293518064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,2,power_law_1.2,0.386297607421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,2,power_law_1.2,0.40353918075561523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,2,power_law_1.2,0.4083392143249512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,2,power_law_1.2,0.4195712089538574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,2,power_law_1.2,0.4395711898803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,2,power_law_1.2,0.46184959411621096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,2,power_law_1.2,0.5101823806762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,2,power_law_1.2,0.593990421295166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,2,power_law_1.2,0.6777791976928711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,2,power_law_1.2,0.8496831893920899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,2,power_law_1.2,0.10185600519180298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,2,power_law_1.2,0.9774016380310059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,2,power_law_1.2,0.09720960259437561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,2,power_law_1.2,1.3890368461608886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,2,power_law_1.2,0.1223680019378662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,2,power_law_1.2,1.6118528366088867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,2,power_law_1.2,0.16782079935073851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,2,power_law_1.2,0.2183680057525635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,2,power_law_1.2,2.290470314025879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,2,power_law_1.2,0.3276416063308716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,2,power_law_1.2,3.113209533691406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,2,power_law_1.2,0.36353919506072996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,2,power_law_1.2,3.8171646118164064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,2,power_law_1.2,0.3795072078704834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,2,power_law_1.2,0.3745471954345703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,2,power_law_1.2,5.552787017822266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,2,power_law_1.2,0.3880000114440918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,2,power_law_1.2,0.3882688045501709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,2,power_law_1.2,10.833612823486328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,2,power_law_1.2,0.40472960472106934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,2,power_law_1.2,0.40015358924865724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,2,power_law_1.2,0.41504640579223634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,2,power_law_1.2,0.42195839881896974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,2,power_law_1.2,0.4149631977081299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,2,power_law_1.2,0.4419839859008789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,2,power_law_1.2,0.4778560161590576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.2,0.518016004562378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.2,0.5926400184631347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.2,0.685811185836792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.2,0.8457856178283691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.2,0.9899968147277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.2,1.2838591575622558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.2,1.6529151916503906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.2,2.311507225036621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.2,2.9252159118652346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,2,power_law_1.2,3.7493568420410157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,2,power_law_1.2,5.669657516479492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,2,power_law_1.2,9.934598541259765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,2,power_law_1.2,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,2,power_law_1.2,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,2,power_law_1.2,0.04751999974250794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,2,power_law_1.2,0.06849279999732971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,2,power_law_1.2,0.08253440260887146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,2,power_law_1.2,0.1371072053909302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,2,power_law_1.2,0.14151040315628052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,2,power_law_1.2,0.14975359439849853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,2,power_law_1.2,0.15791360139846802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,2,power_law_1.2,0.15110399723052978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,2,power_law_1.2,0.16513919830322266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,2,power_law_1.2,0.17224960327148436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,2,power_law_1.2,0.17027839422225952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,2,power_law_1.2,0.18024319410324097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,2,power_law_1.2,0.1859392046928406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,2,power_law_1.2,0.21838080883026123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,2,power_law_1.2,0.1938815951347351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,2,power_law_1.2,0.3939584016799927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.2,0.44004478454589846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.2,0.3238719940185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.2,0.33477120399475097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.2,0.38108799457550047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.2,0.44788479804992676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.2,0.5494976043701172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.2,0.718291187286377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.2,0.9244928359985352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.2,1.1961536407470703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,1,balanced,0.09754133224487305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,2,power_law_1.2,1.4694272041320802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,1,balanced,0.14962133765220642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,1,balanced,0.2205066680908203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,2,power_law_1.2,2.2276416778564454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,1,balanced,0.3290879925092061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,1,balanced,0.5763200124104818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,2,power_law_1.2,4.283404922485351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,1,balanced,0.6772159735361735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,1,balanced,0.6809919675191244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,1,balanced,0.6668533484141032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,1,balanced,0.6715679963429769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,1,balanced,0.6797653039296468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,1,balanced,0.6717706521352133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,1,balanced,0.6815573374430338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,1,balanced,0.6903839906056722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,1,balanced,0.6899946530659994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,1,balanced,0.6945333480834961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,1,balanced,0.7085119883219401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,1,balanced,0.6959253152211508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,1,balanced,0.7464853127797445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,1,balanced,0.7525119781494141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,1,balanced,0.813157320022583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,1,balanced,0.8663093249003092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,1,balanced,0.07339733342329662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,1,balanced,0.10234133402506511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,1,balanced,1.3324906826019287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,1,balanced,0.13961600263913473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,1,balanced,0.20786666870117188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,1,balanced,0.32940266529719037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,1,balanced,0.3883039951324463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,1,balanced,0.3776479959487915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,1,balanced,1.352128028869629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,1,balanced,0.3781546751658122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,1,balanced,0.39266665776570636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,1,balanced,0.39502934614817303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,1,balanced,0.3911093473434448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,1,balanced,0.40358400344848633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,1,balanced,1.9900372823079426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,1,balanced,0.3957120180130005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,1,balanced,0.4106080134709676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,1,balanced,0.4112853209177653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,1,balanced,0.4174186786015828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,1,balanced,0.4328746795654297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,1,balanced,2.6078294118245444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,1,balanced,0.47627735137939453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,1,balanced,0.49965333938598633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,1,balanced,0.573082685470581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,1,balanced,0.6390933195749918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,1,balanced,3.170207977294922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,1,balanced,0.9500373204549154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,1,balanced,1.0523040294647217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,1,balanced,0.09726933638254802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,1,balanced,0.12329066793123881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,1,balanced,0.17839999993642172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,1,balanced,1.528175989786784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,1,balanced,0.2757333318392436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,1,balanced,4.443509419759114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,1,balanced,0.4254026810328166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,1,balanced,0.46804265181223553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,1,balanced,0.4484106699625651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,1,balanced,0.4487786690394084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,1,balanced,0.45585068066914874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,1,balanced,0.4447093407313029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,1,balanced,2.002229372660319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,1,balanced,0.44577598571777344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,1,balanced,0.45555198192596436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,1,balanced,0.46589867273966473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,1,balanced,0.4740746815999349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,1,balanced,0.4604213237762451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,1,balanced,0.47860801219940186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,1,balanced,0.49167998631795246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,1,balanced,4.986261367797852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,1,balanced,2.608309268951416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,1,balanced,0.5189599990844727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,1,balanced,0.5439626773198446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,1,balanced,0.5989760160446167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,1,balanced,0.6630133390426636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,1,balanced,1.0824159781138103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,1,balanced,3.520053227742513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,1,balanced,1.1831626892089844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,1,balanced,0.025034666061401367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,1,balanced,1.6776906649271648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,1,balanced,0.028522667785485584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,1,balanced,0.04244266450405121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,1,balanced,8.039695739746094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,1,balanced,0.05916800101598104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,1,balanced,0.09090133508046468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,1,balanced,0.11148266990979512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,1,balanced,4.126565297444661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,1,balanced,0.1162559986114502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,1,balanced,0.12289599577585857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,1,balanced,2.1921440760294595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,1,balanced,0.1267039974530538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,1,balanced,0.13221333424250284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,1,balanced,0.14341866970062256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,1,balanced,0.15387733777364096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,1,balanced,0.13557333747545877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,1,balanced,0.14521599809328714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,1,balanced,0.1556000014146169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,1,balanced,0.14725333452224731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,1,balanced,0.15546666582425436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,1,balanced,2.8719199498494468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,1,balanced,0.17413866519927979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,1,balanced,0.17961599429448447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,1,balanced,0.3261920015017192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,1,balanced,0.361407995223999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,1,balanced,6.644816080729167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,1,balanced,0.3173706730206807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,1,balanced,3.93122132619222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,1,balanced,0.3285706639289856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,1,balanced,0.534608006477356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,1,balanced,4.606287956237793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,1,balanced,0.5864693323771158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,1,balanced,16.018447875976562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,1,balanced,0.8371733029683431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,1,balanced,13.129178365071615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,1,balanced,7.288368225097656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,1,balanced,1.0862399737040203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,1,balanced,1.184714635213216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,1,balanced,14.341108957926432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,1,balanced,1.957141399383545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,1,power_law_1.01,0.0973695993423462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,1,power_law_1.01,0.12406400442123414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,1,power_law_1.01,0.16837120056152344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,1,power_law_1.01,0.2670655965805054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,1,power_law_1.01,0.3396415948867798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,1,power_law_1.01,0.5251391887664795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,1,power_law_1.01,0.5657919883728028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,1,power_law_1.01,0.5749248027801513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,1,power_law_1.01,0.6009215831756591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,1,power_law_1.01,0.6009344100952149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,1,power_law_1.01,0.6321856021881104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,1,power_law_1.01,0.6404672145843506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,1,power_law_1.01,0.6380224227905273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,1,balanced,3.687525431315104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,1,power_law_1.01,0.072953599691391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,1,power_law_1.01,0.6575232028961182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,1,power_law_1.01,0.08455680012702942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,1,power_law_1.01,0.6848896026611329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,1,power_law_1.01,0.10915199518203736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,1,power_law_1.01,0.6813632011413574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,1,power_law_1.01,0.1612671971321106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,1,power_law_1.01,0.7011583805084228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,1,power_law_1.01,0.20153601169586183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,1,power_law_1.01,0.7778240203857422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,1,power_law_1.01,0.3041984081268311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,1,power_law_1.01,0.8316608428955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,1,power_law_1.01,0.32452480792999266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,1,power_law_1.01,0.950931167602539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,1,power_law_1.01,0.33644158840179444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,1,power_law_1.01,1.0567999839782716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,1,power_law_1.01,0.3476223945617676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,1,power_law_1.01,1.2997056007385255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,1,power_law_1.01,0.34780800342559814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,1,power_law_1.01,1.5467776298522948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,1,power_law_1.01,0.36229119300842283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,1,power_law_1.01,2.0294591903686525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,1,power_law_1.01,0.36545279026031496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,1,power_law_1.01,2.503500747680664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,1,power_law_1.01,0.3824064016342163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,1,power_law_1.01,3.472025680541992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,1,power_law_1.01,0.3873600006103516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,1,power_law_1.01,0.40342397689819337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,1,power_law_1.01,4.3921855926513675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,1,power_law_1.01,0.40724477767944334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,1,power_law_1.01,5.349926376342774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,1,power_law_1.01,0.42695040702819825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,1,power_law_1.01,8.197650909423828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,1,power_law_1.01,0.4871103763580322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,1,power_law_1.01,0.526636791229248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,1,power_law_1.01,0.6241471767425537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,1,power_law_1.01,15.942591857910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,1,power_law_1.01,0.7216512203216553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,1,power_law_1.01,0.9163968086242675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,1,power_law_1.01,1.1101247787475585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,1,power_law_1.01,1.4899264335632325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,1,power_law_1.01,1.873036766052246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,1,power_law_1.01,2.644607925415039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,1,power_law_1.01,3.4265151977539063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,1,power_law_1.01,4.228102493286133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,1,power_law_1.01,6.5646209716796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,1,power_law_1.01,12.887648010253907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,1,power_law_1.01,0.09793919920921326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,1,power_law_1.01,0.10903040170669556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,1,power_law_1.01,0.14489599466323852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,1,power_law_1.01,0.1959488034248352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,1,power_law_1.01,0.24601600170135499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,1,power_law_1.01,0.36195199489593505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,1,power_law_1.01,0.3840320110321045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,1,power_law_1.01,0.39064319133758546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,1,power_law_1.01,0.41573119163513184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,1,power_law_1.01,0.41367039680480955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,1,power_law_1.01,0.4335296154022217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,1,power_law_1.01,0.44161281585693357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,1,power_law_1.01,0.4396992206573486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,1,power_law_1.01,0.4543039798736572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,1,power_law_1.01,0.46364798545837405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,1,power_law_1.01,0.4706431865692139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,1,power_law_1.01,0.4844480037689209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,1,power_law_1.01,0.551968002319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.01,0.5993279933929443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.01,0.7139008045196533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.01,0.8231616020202637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.01,1.0464960098266602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.01,1.2629247665405274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.01,1.6931520462036134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.01,2.130931282043457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.01,2.993657684326172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.01,3.848582458496094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.01,4.700691223144531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,1,power_law_1.01,7.302738952636719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,1,power_law_1.01,14.304261779785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,1,power_law_1.01,0.023520000278949738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,1,power_law_1.01,0.030963200330734252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,1,power_law_1.01,0.03856000006198883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,1,power_law_1.01,0.059443199634552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,1,power_law_1.01,0.08035839796066284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,1,power_law_1.01,0.11674239635467529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,1,power_law_1.01,0.14204800128936768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,1,power_law_1.01,0.14615039825439452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,1,power_law_1.01,0.15327999591827393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,1,power_law_1.01,0.16031999588012696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,1,power_law_1.01,0.17466880083084108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,1,power_law_1.01,0.18676480054855346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,1,power_law_1.01,0.167302405834198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,1,power_law_1.01,0.180511999130249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,1,power_law_1.01,0.18897279500961303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,1,power_law_1.01,0.17703039646148683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,1,power_law_1.01,0.1923583984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,1,power_law_1.01,0.2137984037399292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.01,0.2265023946762085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.01,0.4324672222137451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.01,0.5039743900299072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.01,0.3694911956787109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.01,0.4245440006256104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.01,0.5164415836334229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.01,0.6164735794067383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.01,0.8412608146667481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.01,1.0550975799560547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,1,power_law_1.2,0.09451519846916198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,1,power_law_1.2,0.11224960088729859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.01,1.2875136375427245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,1,power_law_1.2,0.16747519969940186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,1,power_law_1.2,0.07277439832687378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,1,power_law_1.01,1.9498624801635742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,1,power_law_1.2,0.2411776065826416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,1,power_law_1.2,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,1,power_law_1.01,3.7366718292236327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,1,power_law_1.2,0.3010432004928589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,1,power_law_1.2,0.11010559797286987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,1,power_law_1.2,0.5117184162139893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,1,power_law_1.2,0.14897279739379882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,1,power_law_1.2,0.557535982131958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,1,power_law_1.2,0.1821887969970703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,1,power_law_1.2,0.3075200080871582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,1,power_law_1.2,0.5892032146453857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,1,power_law_1.2,0.33244800567626953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,1,power_law_1.2,0.6100863933563232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,1,power_law_1.2,0.3410239934921265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,1,power_law_1.2,0.6420415878295899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,1,power_law_1.2,0.35438721179962157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,1,power_law_1.2,0.6409088134765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,1,power_law_1.2,0.3647423982620239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,1,power_law_1.2,0.6633471965789794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,1,power_law_1.2,0.3728960037231445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,1,power_law_1.2,0.6739647865295411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,1,power_law_1.2,0.3789247989654541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,1,power_law_1.2,0.6843455791473388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,1,power_law_1.2,0.387827205657959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,1,power_law_1.2,0.6886591911315918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,1,power_law_1.2,0.39809279441833495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,1,power_law_1.2,0.7079296112060547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,1,power_law_1.2,0.7254911899566651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,1,power_law_1.2,0.4105088233947754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,1,power_law_1.2,0.8129088401794433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,1,power_law_1.2,0.4224575996398926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,1,power_law_1.2,0.44173440933227537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,1,power_law_1.2,0.8551808357238769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,1,power_law_1.2,0.9669376373291015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,1,power_law_1.2,0.5005055904388428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,1,power_law_1.2,1.0942591667175292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,1,power_law_1.2,1.3407296180725097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,1,power_law_1.2,1.5630592346191405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,1,power_law_1.2,2.0290943145751954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,1,power_law_1.2,2.5109888076782227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,1,power_law_1.2,3.4491649627685548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,1,power_law_1.2,4.389817428588867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,1,power_law_1.2,5.285593414306641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,1,power_law_1.2,8.122822570800782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,1,power_law_1.2,15.662278747558593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,1,power_law_1.2,0.09658240079879761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,1,power_law_1.2,0.10092799663543701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,1,power_law_1.2,0.1365056037902832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,1,power_law_1.2,0.18946559429168702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,1,power_law_1.2,0.23145599365234376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,1,power_law_1.2,0.38056321144104005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,1,power_law_1.2,0.4004096031188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,1,power_law_1.2,0.40607361793518065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,1,power_law_1.2,0.42055039405822753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,1,power_law_1.2,0.4400512218475342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,1,power_law_1.2,0.4463871955871582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,1,power_law_1.2,0.4526976108551025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,1,power_law_1.2,0.45948162078857424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,1,power_law_1.2,0.45710082054138185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,1,power_law_1.2,0.48275198936462405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,1,power_law_1.2,0.47934718132019044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,1,power_law_1.2,0.5099967956542969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,1,power_law_1.2,0.5724927902221679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.2,0.6183872222900391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.2,0.7316736221313477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.2,0.8305536270141601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.2,1.058675193786621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.2,1.274137592315674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.2,1.701228713989258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.2,2.1291967391967774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.2,2.9796415328979493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.2,3.8389183044433595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,1,power_law_1.2,0.023846399784088135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,1,power_law_1.2,0.027110400795936584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.2,4.683200073242188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,1,power_law_1.2,0.03960959911346436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,1,power_law_1.2,7.271711730957032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,1,power_law_1.2,0.055769598484039305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,1,power_law_1.2,0.07624319791793824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,1,power_law_1.2,0.12149759531021118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,1,power_law_1.2,14.231852722167968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,1,power_law_1.2,0.13571840524673462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,1,power_law_1.2,0.14693119525909423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,1,power_law_1.2,0.15286400318145751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,1,power_law_1.2,0.1571328043937683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,1,power_law_1.2,0.17461760044097902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,1,power_law_1.2,0.1864127993583679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,1,power_law_1.2,0.1725823998451233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,1,power_law_1.2,0.18309119939804078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,1,power_law_1.2,0.19697920083999634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,1,power_law_1.2,0.18480639457702636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,1,power_law_1.2,0.1879487991333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,1,power_law_1.2,0.2244863986968994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.2,0.22914559841156007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.2,0.4350143909454346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.2,0.5077951908111572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.2,0.36501760482788087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.2,0.4211584091186523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.2,0.5165696144104004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.2,0.6307519912719727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.2,0.8510592460632325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.2,1.0723648071289062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.2,1.2935935974121093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,1,power_law_1.2,1.9664447784423829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,1,power_law_1.2,3.7447681427001953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,4,balanced,0.09653866291046143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,4,balanced,0.12949333588282266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,1,power_law_1.2,0.5403647899627686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,4,balanced,0.18918399016062418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,4,balanced,0.31868799527486164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,1,power_law_1.2,0.6403840065002442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,4,balanced,0.5571253299713135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,1,power_law_1.2,0.7291327953338623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,4,balanced,0.9952320257822672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,1,power_law_1.2,0.9271488189697266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,4,balanced,1.4460320472717285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,1,power_law_1.2,1.1220159530639648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,4,balanced,1.4488159815470378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,4,balanced,1.441520055135091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,1,power_law_1.2,1.5073151588439941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,4,balanced,1.46725861231486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,1,power_law_1.2,1.8780223846435546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,4,balanced,1.511306603749593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,1,power_law_1.2,2.642188835144043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,4,balanced,1.4801279703776042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,4,balanced,1.4961546262105305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,1,power_law_1.2,3.4149185180664063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,4,balanced,1.456293265024821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,1,power_law_1.2,4.198271942138672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,4,balanced,1.4995519320170085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,1,power_law_1.2,6.550611114501953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,4,balanced,1.438528060913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,4,balanced,1.489829381306966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,1,power_law_1.2,12.785260772705078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,4,balanced,1.4418187141418457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,4,balanced,1.4902559916178386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,4,balanced,1.489535967508952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,4,balanced,1.5365440050760906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,4,balanced,1.5201066335042317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,4,balanced,1.555221398671468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,4,balanced,1.6633812586466472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,4,balanced,2.7347199122111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,4,balanced,0.07070933282375336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,4,balanced,0.08873599767684937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,4,balanced,0.12873599926630655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,4,balanced,0.19871999820073447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,4,balanced,2.5076160430908203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,4,balanced,0.3279680013656616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,4,balanced,0.551530679066976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,4,balanced,0.7866079807281494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,4,balanced,0.8071200052897135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,4,balanced,0.7678986390431722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,4,balanced,0.7636906305948893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,4,balanced,0.7867680390675863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,4,balanced,0.7732746601104736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,4,balanced,0.798805316289266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,4,balanced,0.8014132976531982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,4,balanced,0.7994720141092936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,4,balanced,0.8231786886850992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,4,balanced,0.8332053025563558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,4,balanced,3.668053309122721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,4,balanced,0.8091626962025961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,4,balanced,0.835530678431193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,4,balanced,0.8506666819254557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,4,balanced,0.8840533097585043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,4,balanced,0.0788213312625885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,4,balanced,0.10068266590436299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,4,balanced,0.9292960166931152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,4,balanced,0.1528426706790924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,4,balanced,0.2413333257039388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,4,balanced,0.33560001850128174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,4,balanced,0.5962346792221069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,4,balanced,0.989349365234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,4,balanced,0.7813386917114258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,4,balanced,0.7855679988861084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,4,balanced,0.7637813091278076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,4,balanced,5.138976097106934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,4,balanced,0.7457653681437174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,4,balanced,0.7333172957102457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,4,balanced,0.7541493574778239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,4,balanced,1.1360320250193279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,4,balanced,0.7258933385213217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,4,balanced,0.7332320213317871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,4,balanced,0.7662506898244222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,4,balanced,0.7775946458180746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,4,balanced,0.7479306856791178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,4,balanced,1.7224052747090657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,4,balanced,0.7829333146413168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,4,balanced,0.782533327738444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,4,balanced,0.8283893267313639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,4,balanced,0.8193333148956299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,4,balanced,1.8058026631673176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,4,balanced,0.8505600293477377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,4,balanced,0.8525226910909017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,4,power_law_1.01,0.1941248059272766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,4,balanced,7.404799779256185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,4,power_law_1.01,0.33500161170959475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,4,power_law_1.01,0.5622719764709473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,4,power_law_1.01,0.26162559986114503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,4,balanced,0.9695680141448975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,4,power_law_1.01,0.4100543975830078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,4,balanced,2.526533285776774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,4,power_law_1.01,0.6489151954650879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,4,power_law_1.01,0.7346176147460938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,4,power_law_1.01,0.9904512405395508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,4,balanced,1.623045285542806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,4,power_law_1.01,1.0280960083007813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,4,power_law_1.01,0.9789055824279785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,4,power_law_1.01,1.0361984252929688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,4,power_law_1.01,1.143667221069336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,4,power_law_1.01,1.1943296432495116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,4,power_law_1.01,1.1495679855346679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,4,power_law_1.01,1.2031871795654296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,4,balanced,1.6353599230448406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,4,power_law_1.01,1.1474431991577148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,4,balanced,3.436469395955404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,4,power_law_1.01,1.2112064361572266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,4,power_law_1.01,1.2419903755187989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,4,power_law_1.01,1.3312959671020508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,4,power_law_1.01,1.4056256294250489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,4,power_law_1.01,1.5653951644897461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,4,power_law_1.01,1.8341056823730468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,4,balanced,2.3390986124674478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,4,power_law_1.01,2.1048511505126952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,4,power_law_1.01,2.4251071929931642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,4,power_law_1.01,2.946323204040527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,4,power_law_1.01,3.7490943908691405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,4,power_law_1.01,4.635878372192383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,4,power_law_1.01,5.428979110717774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,4,power_law_1.01,7.990322875976562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,4,balanced,5.045951843261719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,4,power_law_1.01,15.359027099609374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,4,balanced,3.1845973332722983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,4,balanced,12.592458089192709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,4,balanced,4.6897226969401045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,4,balanced,8.807519912719727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,4,balanced,8.129823684692383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,4,power_law_1.01,0.12915199995040894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,4,power_law_1.01,0.19781119823455812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,4,power_law_1.01,0.31560320854187013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,4,power_law_1.01,0.13815040588378907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,4,power_law_1.01,0.17734400033950806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,4,power_law_1.01,0.2097856044769287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,4,power_law_1.01,0.2488640069961548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,4,power_law_1.01,0.3428992033004761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,4,power_law_1.01,0.36788480281829833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,4,power_law_1.01,0.1882240056991577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,4,power_law_1.01,0.4268352031707764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,4,power_law_1.01,0.27206399440765383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,4,power_law_1.01,0.5424384117126465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,4,power_law_1.01,0.40203518867492677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,4,power_law_1.01,0.5485631942749023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,4,power_law_1.01,0.4395904064178467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,4,power_law_1.01,0.574944019317627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,4,power_law_1.01,0.5567935943603516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,4,power_law_1.01,0.6043968200683594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,4,power_law_1.01,0.5550271987915039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,4,power_law_1.01,0.6387775897979736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,4,power_law_1.01,0.5712704181671142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,4,power_law_1.01,0.6153024196624756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,4,power_law_1.01,0.5992447853088378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,4,power_law_1.01,0.6454336166381835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,4,power_law_1.01,0.6105984210968017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,4,power_law_1.01,0.6621952056884766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,4,power_law_1.01,0.6574912071228027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,4,power_law_1.01,0.680940818786621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,4,power_law_1.01,0.6542975902557373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,4,power_law_1.01,0.678764820098877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,4,power_law_1.01,0.656390380859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,4,power_law_1.01,0.7141439914703369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,4,power_law_1.01,0.6673535823822021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,4,power_law_1.01,0.7578368186950684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,4,power_law_1.01,0.6897984027862549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,4,power_law_1.01,0.8417280197143555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,4,power_law_1.01,0.715999984741211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,4,power_law_1.01,0.711840009689331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,4,power_law_1.01,0.8027584075927734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,4,power_law_1.01,0.9081151962280274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,4,power_law_1.01,1.026912021636963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,4,power_law_1.01,1.1849023818969726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,4,power_law_1.01,1.516320037841797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,4,power_law_1.01,1.7728191375732423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,4,power_law_1.01,2.3497407913208006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,4,power_law_1.2,0.19728000164031984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,4,power_law_1.01,2.9458944320678713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,4,power_law_1.2,0.3038912057876587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,4,power_law_1.01,3.5314815521240233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,4,power_law_1.2,0.3491199970245361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,4,power_law_1.01,5.077248001098633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,4,power_law_1.2,0.296287989616394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,4,power_law_1.2,0.40862717628479006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,4,power_law_1.01,9.432147216796874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,4,power_law_1.2,0.6079103946685791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,4,power_law_1.2,0.7058815956115723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,4,power_law_1.2,0.9363391876220704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,4,power_law_1.2,0.9584959983825684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,4,power_law_1.2,0.1398848056793213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,4,power_law_1.2,0.9827648162841797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,4,power_law_1.2,0.20510079860687255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,4,power_law_1.2,1.0531647682189942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,4,power_law_1.2,0.22168960571289062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,4,power_law_1.2,1.1459391593933106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,4,power_law_1.2,0.17095040082931517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,4,power_law_1.2,1.1954367637634278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,4,power_law_1.2,0.2554944038391113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,4,power_law_1.2,1.1492032051086425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,4,power_law_1.2,0.35637118816375735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,4,power_law_1.2,1.209273624420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,4,power_law_1.2,0.3855936050415039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,4,power_law_1.2,1.2138879776000977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,4,power_law_1.2,0.544697618484497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,4,power_law_1.2,1.232819175720215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,4,power_law_1.2,0.5499392032623291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,4,power_law_1.2,1.3559040069580077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,4,power_law_1.2,0.5677567958831787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,4,power_law_1.2,1.3820672035217285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,4,power_law_1.2,0.6159872055053711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,4,power_law_1.2,1.6047807693481446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,4,power_law_1.2,0.5933440208435059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,4,power_law_1.2,1.7621055603027345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,4,power_law_1.2,0.6655295848846435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,4,power_law_1.2,1.9847232818603515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,4,power_law_1.2,0.6615424156188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,4,power_law_1.2,2.2764799118041994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,4,power_law_1.2,0.6950399875640869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,4,power_law_1.2,2.5839168548583986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,4,power_law_1.2,0.6690048217773438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,4,power_law_1.2,3.049056053161621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,4,power_law_1.2,0.723360013961792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,4,power_law_1.2,3.980953598022461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,4,power_law_1.2,0.763865613937378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,4,power_law_1.2,5.1382080078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,4,power_law_1.2,0.8037952423095703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,4,power_law_1.2,6.392531204223633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,4,power_law_1.2,0.9156096458435059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,4,power_law_1.2,1.0190848350524901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,4,power_law_1.2,9.650259399414063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,4,power_law_1.2,1.1796031951904298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,4,power_law_1.2,1.3491071701049804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,4,power_law_1.2,16.33811798095703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,4,power_law_1.2,1.7145984649658204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,4,power_law_1.2,1.903968048095703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,4,power_law_1.2,2.6303808212280275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,4,power_law_1.2,3.392243194580078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,4,power_law_1.2,4.111091232299804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,4,power_law_1.2,6.028761672973633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,4,power_law_1.2,11.019123077392578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,4,power_law_1.2,0.13770240545272827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,4,power_law_1.01,0.9251839637756347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,4,power_law_1.2,0.21614720821380615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,4,power_law_1.01,1.0531135559082032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,4,power_law_1.2,0.18268799781799316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,4,power_law_1.01,1.237945556640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,4,power_law_1.2,0.16872960329055786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,4,power_law_1.01,1.5783488273620605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,4,power_law_1.2,0.24931199550628663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,4,power_law_1.01,1.8569023132324218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,4,power_law_1.2,0.3506752014160156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,4,power_law_1.01,2.4137088775634767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,4,power_law_1.2,0.3873600006103516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,4,power_law_1.01,3.0386688232421877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,4,power_law_1.2,0.523801612854004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,4,power_law_1.01,3.605587387084961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,4,power_law_1.2,0.5434944152832031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,4,power_law_1.2,0.5432447910308837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,4,power_law_1.01,5.618815994262695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,4,power_law_1.2,0.6060736179351807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,4,power_law_1.01,10.480908966064453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,4,power_law_1.2,0.6302464008331299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,4,power_law_1.2,0.6383808135986329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,4,power_law_1.2,0.6554175853729248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,4,power_law_1.2,0.6778431892395019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,4,power_law_1.2,0.6655807971954346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,4,power_law_1.2,0.7172800064086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,4,power_law_1.2,0.7299967765808105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,4,power_law_1.2,0.7844736099243164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,4,power_law_1.2,0.8544575691223144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,4,power_law_1.2,0.9623040199279785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,4,power_law_1.2,1.1248767852783204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,4,power_law_1.2,1.2928064346313477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,4,power_law_1.2,1.5962944030761719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,4,power_law_1.2,2.042470359802246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,4,power_law_1.2,2.5938047409057616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,4,power_law_1.2,3.3125377655029298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,4,power_law_1.2,3.6923583984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,2,balanced,0.10086933771769206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,2,balanced,0.1256106694539388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,2,balanced,0.19087467590967813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,4,power_law_1.2,5.575769424438477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,2,balanced,0.33210132519404095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,2,balanced,0.5473119815190634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,4,power_law_1.2,10.921734619140626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,2,balanced,0.967093308766683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,2,balanced,1.3869120279947917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,2,balanced,1.419914722442627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,2,balanced,1.3959040641784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,2,balanced,1.4083520571390789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,2,balanced,1.4372533162434895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,2,balanced,1.4375519752502441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,2,balanced,0.075013334552447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,2,balanced,1.3894720077514648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,2,balanced,0.08980799714724223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,2,balanced,1.422826608022054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,2,balanced,0.12335466345151265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,2,balanced,1.4349652926127117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,2,balanced,0.20198933283487955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,2,balanced,0.3279413382212321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,2,balanced,1.472117265065511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,2,balanced,0.5795360008875529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,2,balanced,1.450229326883952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,2,balanced,0.782474676767985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,2,balanced,0.7922080357869467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,2,balanced,1.4189119338989258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,2,balanced,0.779269297917684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,2,balanced,0.7530186971028646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,2,balanced,1.488111972808838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,2,balanced,0.7575733661651611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,2,balanced,0.7539467016855875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,2,balanced,1.4854507446289062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,2,balanced,0.7571626504262289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,2,balanced,0.7770613034566244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,2,balanced,0.7862293720245361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,2,balanced,1.5306240717569988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,2,balanced,0.7955093383789062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,2,balanced,0.7943466504414877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,2,balanced,1.5918080012003581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,2,balanced,0.808672030766805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,2,balanced,0.8322133223215739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,2,balanced,0.8813546498616537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,2,balanced,1.6585973103841145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,2,balanced,0.9305493036905924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,2,balanced,1.0155733426411946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,2,balanced,1.8335679372151692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,2,balanced,1.1195680300394695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,2,balanced,2.7797012329101562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,2,balanced,1.335840066274007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,2,balanced,1.9149173100789387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,2,balanced,2.8048054377237954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,2,balanced,2.2022933959960938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,2,balanced,0.08935466408729553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,2,balanced,0.10656000177065532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,2,balanced,0.15610667069753012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,2,balanced,0.21788267294565836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,2,balanced,0.37246398131052655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,2,balanced,0.5620319843292236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,2,balanced,0.8046613534291586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,2,balanced,0.7988320191701254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,2,balanced,0.7821013132731119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,2,balanced,3.9646078745524087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,2,balanced,0.770085334777832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,2,balanced,2.97272523244222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,2,balanced,0.7588640054066976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,2,balanced,0.7451732953389486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,2,balanced,0.7280320326487223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,2,balanced,0.7377706368764242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,2,balanced,0.7803893089294434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,2,balanced,0.7568053404490153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,2,balanced,0.7729012966156006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,2,balanced,0.7921013037363688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,2,balanced,0.7819999853769938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,2,balanced,0.8421386877695719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,2,balanced,3.8935254414876304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,2,balanced,0.8860639731089274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,2,balanced,5.430986404418945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,2,balanced,0.9523999691009521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,2,balanced,1.0073333581288655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,2,power_law_1.01,0.12854399681091308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,2,power_law_1.01,0.1978368043899536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,2,power_law_1.01,0.3244607925415039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,2,balanced,1.2309760252634685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,2,power_law_1.01,0.2510528087615967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,2,power_law_1.01,0.43715839385986327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,2,power_law_1.01,0.5922560214996337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,2,power_law_1.01,0.6935808181762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,2,balanced,1.8740426699320476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,2,balanced,5.903770446777344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,2,power_law_1.01,0.9352383613586426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,2,power_law_1.01,0.9279168128967286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,2,power_law_1.01,0.9945023536682129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,2,power_law_1.01,1.04268798828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,2,power_law_1.01,1.071571159362793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,2,balanced,7.9534454345703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,2,power_law_1.01,1.0887167930603028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,2,balanced,2.118000030517578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,2,power_law_1.01,1.1656448364257812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,2,power_law_1.01,1.1411199569702148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,2,power_law_1.01,1.1947392463684081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,2,power_law_1.01,1.2156607627868652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,2,power_law_1.01,1.2433792114257813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,2,power_law_1.01,1.3164159774780273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,2,power_law_1.01,1.4446592330932617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,2,power_law_1.01,1.5453120231628419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,2,balanced,2.9381281534830728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,2,power_law_1.01,1.7871936798095702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,2,power_law_1.01,2.0181247711181642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,2,power_law_1.01,2.529484748840332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,2,power_law_1.01,2.942643165588379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,2,power_law_1.01,3.841574478149414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,2,power_law_1.01,4.825420761108399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,2,power_law_1.01,5.684447860717773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,2,power_law_1.01,8.477401733398438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,2,balanced,3.8190720876057944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,2,power_law_1.01,15.632537841796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,2,balanced,10.827140808105469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,2,balanced,5.728000005086263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,2,balanced,14.363253275553385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,2,balanced,10.443536122639975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,2,power_law_1.01,0.09208319783210754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,2,power_law_1.01,0.12828160524368287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,2,power_law_1.01,0.2030911922454834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,2,power_law_1.01,0.11945600509643554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,2,power_law_1.01,0.16224639415740966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,2,power_law_1.01,0.1537343978881836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,2,power_law_1.01,0.25383040904998777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,2,power_law_1.01,0.22288000583648682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,2,power_law_1.01,0.3523328065872192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,2,power_law_1.01,0.16061439514160156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,2,power_law_1.01,0.4173120021820068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,2,power_law_1.01,0.2785664081573486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,2,power_law_1.01,0.533465576171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,2,power_law_1.01,0.412659215927124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,2,power_law_1.01,0.5444736003875732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,2,power_law_1.01,0.4510015964508057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,2,power_law_1.01,0.5558207988739013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,2,power_law_1.01,0.5380671977996826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,2,power_law_1.01,0.5813632011413574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,2,power_law_1.01,0.5713727951049805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,2,power_law_1.01,0.6085440158843994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,2,power_law_1.01,0.6013311862945556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,2,power_law_1.01,0.5945600032806396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,2,power_law_1.01,0.6052224159240722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,2,power_law_1.01,0.6323584079742431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,2,power_law_1.01,0.6611775875091552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,2,power_law_1.2,0.12553600072860718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,2,power_law_1.01,0.6683072090148926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,2,power_law_1.2,0.19289599657058715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,2,power_law_1.01,0.6726016044616699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,2,power_law_1.2,0.19145599603652955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,2,power_law_1.01,0.6969279766082763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,2,power_law_1.2,0.2594624042510986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,2,power_law_1.01,0.7138432025909424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,2,power_law_1.2,0.4113920211791992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,2,power_law_1.01,0.7732992172241211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,2,power_law_1.2,0.5570496082305908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,2,power_law_1.01,0.8333439826965332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,2,power_law_1.2,0.6180736064910889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,2,power_law_1.01,0.921395206451416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,2,power_law_1.2,0.8690752029418946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,2,power_law_1.01,1.1298432350158691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,2,power_law_1.2,0.9387328147888183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,2,power_law_1.01,1.2582079887390136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,2,power_law_1.2,0.930016040802002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,2,power_law_1.01,1.6092416763305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,2,power_law_1.2,0.9743552207946777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,2,power_law_1.2,1.0633407592773438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,2,power_law_1.01,1.9480831146240234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,2,power_law_1.2,1.0774208068847657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,2,power_law_1.01,2.6106624603271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,2,power_law_1.2,1.1140480041503906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,2,power_law_1.01,3.3436737060546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,2,power_law_1.2,1.145638370513916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,2,power_law_1.01,3.9828990936279296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,2,power_law_1.2,1.1821375846862794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,2,power_law_1.01,5.879494476318359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,2,power_law_1.2,1.2199999809265136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,2,power_law_1.2,1.3079744338989259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,2,power_law_1.01,11.182157135009765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,2,power_law_1.2,1.3528767585754395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,2,power_law_1.2,1.490604782104492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,2,power_law_1.2,1.6430335998535157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,2,power_law_1.2,1.8808704376220704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,2,power_law_1.2,2.117184066772461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,2,power_law_1.2,2.6617855072021483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,2,power_law_1.2,0.0930176019668579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,2,power_law_1.2,3.218368148803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,2,power_law_1.2,0.12811520099639892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,2,power_law_1.2,4.104377746582031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,2,power_law_1.2,0.11786240339279175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,2,power_law_1.2,5.033919906616211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,2,power_law_1.2,0.14915839433670045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,2,power_law_1.2,0.2552128076553345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,2,power_law_1.2,5.796319961547852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,2,power_law_1.2,0.3271039962768555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,2,power_law_1.2,8.865280151367188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,2,power_law_1.2,0.36808960437774657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,2,power_law_1.2,0.4812032222747803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,2,power_law_1.2,15.988992309570312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,2,power_law_1.2,0.5166272163391114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,2,power_law_1.2,0.5292799949645997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,2,power_law_1.2,0.5694464206695556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,2,power_law_1.2,0.5977344036102294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,2,power_law_1.2,0.6151103973388672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,2,power_law_1.2,0.6068031787872314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,2,power_law_1.2,0.6574272155761719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,2,power_law_1.2,0.6823616027832031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,2,power_law_1.2,0.6928383827209472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,2,power_law_1.2,0.7455872058868408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,2,power_law_1.2,0.7995967864990234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,2,power_law_1.2,0.9130816459655762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,2,power_law_1.2,0.9682175636291503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,2,power_law_1.2,1.1822591781616212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,2,power_law_1.2,1.355635166168213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,2,power_law_1.2,1.735513687133789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,2,power_law_1.2,2.012716865539551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,2,power_law_1.2,2.7454912185668947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,2,power_law_1.2,3.5714561462402346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,2,power_law_1.2,4.058617782592774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,2,power_law_1.2,6.531884765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,2,power_law_1.2,12.081139373779298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,2,power_law_1.01,0.6198400020599365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,2,power_law_1.01,0.6258687973022461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,2,power_law_1.01,0.6567488193511963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,2,power_law_1.01,0.6595583915710449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,2,power_law_1.01,0.6905344009399415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,2,power_law_1.01,0.7420544147491455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,2,power_law_1.01,0.7750720024108887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,2,power_law_1.01,0.839571189880371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,2,power_law_1.2,0.10780160427093506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,2,power_law_1.01,0.9378815650939941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,2,power_law_1.2,0.1533504009246826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,2,power_law_1.01,1.1372480392456055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,2,power_law_1.2,0.14532480239868165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,2,power_law_1.01,1.3036479949951172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,2,power_law_1.2,0.17087359428405763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,2,power_law_1.2,0.2750976085662842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,2,power_law_1.01,1.660211181640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,2,power_law_1.2,0.3500096082687378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,2,power_law_1.01,1.993235206604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,2,power_law_1.2,0.39699199199676516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,2,power_law_1.01,2.6819456100463865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,2,power_law_1.2,0.5310207843780518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,2,power_law_1.01,3.387251281738281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,2,power_law_1.2,0.5448768138885498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,2,power_law_1.01,3.998982238769531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,2,power_law_1.2,0.5559296131134033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,2,power_law_1.2,0.5930240154266357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,2,power_law_1.01,6.117350387573242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,2,power_law_1.2,0.6061888217926026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,2,power_law_1.01,11.91945571899414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,2,power_law_1.2,0.6137792110443115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,2,power_law_1.2,0.6473343849182129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,2,power_law_1.2,0.6679808139801026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,2,power_law_1.2,0.6862592220306396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,2,power_law_1.2,0.708729600906372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,2,power_law_1.2,0.7478208065032959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,1,balanced,0.10806933045387268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,2,power_law_1.2,0.807532787322998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,1,balanced,0.1290773351987203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,2,power_law_1.2,0.8748160362243652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,1,balanced,0.1962239940961202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,1,balanced,0.31612799564997357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,2,power_law_1.2,0.9955583572387695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,1,balanced,0.5361280043919882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,2,power_law_1.2,1.1919168472290038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,1,balanced,0.9780320326487223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,1,balanced,1.3972214063008626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,2,power_law_1.2,1.3558336257934571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,1,balanced,1.4077280362447102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,1,balanced,1.4189866383870442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,2,power_law_1.2,1.682124710083008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,1,balanced,1.3880213101704915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,2,power_law_1.2,2.126393508911133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,1,balanced,1.4076053301493328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,1,balanced,1.4210079511006672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,2,power_law_1.2,2.755072021484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,1,balanced,1.3896373112996419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,2,power_law_1.2,3.4461441040039062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,1,balanced,1.4036906560262044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,1,balanced,1.4142187436421711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,2,power_law_1.2,4.099731063842773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,1,balanced,1.4250772794087727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,2,power_law_1.2,6.496717071533203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,1,balanced,1.4285173416137695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,1,balanced,1.4876213073730469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,2,power_law_1.2,11.737561798095703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,1,balanced,1.5108106931050618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,1,balanced,1.5735467274983723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,1,balanced,1.6269493103027344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,1,balanced,1.7663733164469402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,1,balanced,0.08467732866605122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,1,balanced,0.10003200173377991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,1,balanced,1.906549294789632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,1,balanced,0.13672533631324768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,1,balanced,0.2102666695912679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,1,balanced,0.3250240087509155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,1,balanced,0.562122662862142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,1,balanced,0.7938346862792969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,1,balanced,0.7984480063120524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,1,balanced,2.215226650238037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,1,balanced,0.7958559989929199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,1,balanced,0.7820639610290527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,1,balanced,0.7910186449686686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,1,balanced,0.7954400380452474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,1,balanced,0.8008800347646078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,1,balanced,0.8070293267567953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,1,balanced,0.8268213272094727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,1,balanced,3.1809120178222656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,1,balanced,0.8298772970835367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,1,balanced,0.8479359944661459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,1,balanced,0.8962133725484213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,1,balanced,0.9244106610616049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,1,balanced,1.0091946919759114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,1,balanced,3.5703627268473306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,1,balanced,1.0797119935353596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,1,balanced,1.2453386783599854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,1,balanced,1.4057067235310872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,1,balanced,1.7347040176391602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,1,balanced,4.863183975219727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,1,balanced,2.418501377105713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,1,balanced,3.0440266927083335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,1,balanced,6.370245615641276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,1,balanced,0.12184533476829529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,1,balanced,0.14152533809343973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,1,balanced,0.18599466482798258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,1,balanced,0.2754720052083333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,1,balanced,4.114879926045735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,1,balanced,0.419648011525472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,1,balanced,0.7146186828613281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,1,balanced,0.9096319675445557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,1,balanced,0.9183893203735352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,1,balanced,0.9151466687520345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,1,balanced,0.8887786865234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,1,balanced,0.9019947052001953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,1,balanced,0.884943962097168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,1,balanced,0.8819413185119629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,1,balanced,0.910965363184611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,1,balanced,0.9398826758066813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,1,balanced,0.9165120124816895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,1,balanced,0.959007978439331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,1,balanced,5.182597478230794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,1,balanced,9.579296112060547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,1,balanced,0.9749173323313395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,1,power_law_1.01,0.09821439981460571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,1,balanced,1.0125386714935303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,1,power_law_1.01,0.1260480046272278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,1,power_law_1.01,0.20100479125976561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,1,balanced,1.0676000118255615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,1,power_law_1.01,0.25295999050140383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,1,power_law_1.01,0.40812158584594727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,1,balanced,1.147488037745158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,1,power_law_1.01,0.5936255931854248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,1,power_law_1.01,0.6864384174346924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,1,balanced,1.264965295791626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,1,power_law_1.01,0.8902079582214355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,1,power_law_1.01,0.9220095634460449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,1,power_law_1.01,0.964851188659668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,1,balanced,1.412058671315511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,1,power_law_1.01,1.0145471572875977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,1,power_law_1.01,1.058847999572754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,1,power_law_1.01,1.0659584045410155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,1,balanced,7.87716801961263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,1,power_law_1.01,1.1025152206420898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,1,power_law_1.01,1.138265609741211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,1,balanced,1.7197759946187336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,1,power_law_1.01,1.1688192367553711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,1,power_law_1.01,1.2163200378417969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,1,power_law_1.01,1.268115234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,1,power_law_1.01,1.3352064132690429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,1,power_law_1.01,1.4682815551757813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,1,balanced,2.6115147272745767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,1,power_law_1.01,1.5822272300720215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,1,power_law_1.01,1.8414464950561524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,1,power_law_1.01,2.0933439254760744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,1,power_law_1.01,2.6412479400634767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,1,power_law_1.01,3.186297607421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,1,power_law_1.01,4.228940963745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,1,power_law_1.01,5.289113616943359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,1,balanced,3.1060959498087564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,1,power_law_1.01,6.354816055297851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,1,balanced,17.136576334635418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,1,power_law_1.01,9.437792205810547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,1,power_law_1.01,17.619424438476564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,1,balanced,4.27786127726237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,1,balanced,14.984111785888672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,1,balanced,5.523866653442383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,1,balanced,8.3187255859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,1,power_law_1.01,0.11489280462265014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,1,power_law_1.01,0.07559679746627808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,1,power_law_1.01,0.13911039829254152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,1,power_law_1.01,0.09544960260391236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,1,power_law_1.01,0.18296960592269898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,1,power_law_1.01,0.21895039081573486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,1,power_law_1.01,0.13694080114364623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,1,power_law_1.01,0.34428160190582274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,1,power_law_1.01,0.17004799842834473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,1,power_law_1.01,0.4312895774841309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,1,power_law_1.01,0.25319681167602537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,1,power_law_1.01,0.505075216293335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,1,power_law_1.01,0.35533440113067627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,1,power_law_1.01,0.6229695796966552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,1,power_law_1.01,0.40372481346130373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,1,balanced,15.690218607584635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,1,power_law_1.01,0.6475903987884521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,1,power_law_1.01,0.5288191795349121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,1,power_law_1.01,0.6659904003143311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,1,power_law_1.01,0.5502079963684082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,1,power_law_1.01,0.6920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,1,power_law_1.01,0.5598783969879151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,1,power_law_1.01,0.7083903789520264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,1,power_law_1.01,0.594323205947876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,1,power_law_1.2,0.09733120203018189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,1,power_law_1.01,0.7168255805969238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,1,power_law_1.2,0.1266816020011902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,1,power_law_1.01,0.6210624217987061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,1,power_law_1.01,0.7579967975616455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,1,power_law_1.2,0.16465920209884644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,1,power_law_1.01,0.639244794845581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,1,power_law_1.2,0.23077120780944824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,1,power_law_1.01,0.7576191902160645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,1,power_law_1.01,0.6464191913604737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,1,power_law_1.01,0.7771584033966065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,1,power_law_1.2,0.3624576091766357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,1,power_law_1.01,0.6575104236602783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,1,power_law_1.01,0.8238207817077636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,1,power_law_1.2,0.5073791980743408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,1,power_law_1.01,0.6967040061950683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,1,power_law_1.01,0.8534591674804688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,1,power_law_1.2,0.5842688083648682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,1,power_law_1.01,0.7251008033752442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.01,0.9076607704162598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,1,power_law_1.2,0.8322303771972657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,1,power_law_1.01,0.7793856143951416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,1,power_law_1.2,0.8759360313415527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.01,1.0437376022338867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,1,power_law_1.01,0.8355775833129883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.01,1.152518367767334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,1,power_law_1.01,0.9510144233703614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,1,power_law_1.2,0.9066816329956054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.01,1.386086368560791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,1,power_law_1.2,0.9845376014709473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,1,power_law_1.01,1.0491583824157715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,1,power_law_1.2,1.0142271995544434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.01,1.613657569885254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,1,power_law_1.01,1.2763263702392578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,1,power_law_1.2,1.0283583641052245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.01,2.094003105163574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,1,power_law_1.01,1.483846378326416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,1,power_law_1.2,1.0935359954833985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,1,power_law_1.01,2.555628776550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,1,power_law_1.01,1.9432575225830078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,1,power_law_1.2,1.1152192115783692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,1,power_law_1.01,2.3753984451293944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,1,power_law_1.01,3.489516830444336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,1,power_law_1.2,1.1416128158569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,1,power_law_1.01,3.2443775177001952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,1,power_law_1.01,4.406803131103516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,1,power_law_1.2,1.2041728019714355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,1,power_law_1.01,4.126911926269531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,1,power_law_1.01,5.3307136535644535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,1,power_law_1.2,1.3279744148254395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,1,power_law_1.01,4.9948478698730465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,1,power_law_1.2,1.3696000099182128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,1,power_law_1.01,8.093804931640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,1,power_law_1.2,1.537388801574707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,1,power_law_1.01,7.6264190673828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,1,power_law_1.01,15.451251220703124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,1,power_law_1.2,1.6328319549560546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,1,power_law_1.2,1.9556352615356445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,1,power_law_1.01,14.752088928222657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,1,power_law_1.2,2.2040319442749023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,1,power_law_1.2,2.7127424240112306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,1,power_law_1.2,3.2555774688720702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,1,power_law_1.2,4.325126266479492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,1,power_law_1.2,5.3272449493408205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,1,power_law_1.2,6.342835235595703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,1,power_law_1.2,9.462355041503907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,1,power_law_1.2,17.57021484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,1,power_law_1.2,0.07752320170402527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,8,balanced,0.1536746621131897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,8,balanced,0.1657919983069102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,1,power_law_1.2,0.09815040230751038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,8,balanced,0.16740800937016806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,8,balanced,0.16721065839131674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,1,power_law_1.2,0.11834880113601684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,8,balanced,0.16286933422088623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,1,power_law_1.2,0.1666111946105957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,8,balanced,0.1630293329556783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,8,balanced,0.16216533382733664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,1,power_law_1.2,0.2559936046600342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,8,balanced,0.1590986649195353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,8,balanced,0.15177067120869955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,1,power_law_1.2,0.319488000869751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,8,balanced,0.15286399920781454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,1,power_law_1.2,0.35992319583892823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,8,balanced,0.1476746698220571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,8,balanced,0.1479573349157969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,1,power_law_1.2,0.4946239948272705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,8,balanced,0.14426133036613464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,8,balanced,0.14299199978510538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,1,power_law_1.2,0.5206016063690185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,8,balanced,0.13985066612561545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,1,power_law_1.2,0.547046422958374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,8,balanced,0.13475733002026877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,8,balanced,0.13427733381589255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,1,power_law_1.2,0.5624959945678711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,8,balanced,0.18241065740585327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,1,power_law_1.2,0.5974656105041504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,8,balanced,0.09641599655151367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,8,balanced,0.17072532574335733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,8,balanced,0.09758399923642476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,1,power_law_1.2,0.6097472190856934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,8,balanced,0.21692800521850586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,8,balanced,0.10912533601125081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,8,balanced,0.11384532848993938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,1,power_law_1.2,0.6351168155670166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,1,power_law_1.2,0.11483520269393921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,8,balanced,0.2523519992828369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,8,balanced,0.1011786659558614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,1,power_law_1.2,0.6645440101623535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,8,balanced,0.09983467062314351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,1,power_law_1.2,0.1415295958518982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,8,balanced,0.37834668159484863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,8,balanced,0.10223999619483948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,1,power_law_1.2,0.15777280330657958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,1,power_law_1.2,0.700716781616211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,8,balanced,0.09893332918485005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,8,balanced,0.4605706532796224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,8,balanced,0.10215999682744344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,1,power_law_1.2,0.7292096138000488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,1,power_law_1.2,0.2235327959060669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,8,balanced,0.10007466872533162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,1,power_law_1.2,0.8035584449768066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,1,power_law_1.2,0.30210559368133544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,8,balanced,0.6642560164133707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,8,balanced,0.10365866621335347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,8,balanced,0.09649599591890971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,1,power_law_1.2,0.8531007766723633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,1,power_law_1.2,0.38769919872283937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,8,balanced,0.0981760025024414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,8,balanced,0.8688373565673828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,1,power_law_1.2,1.0118783950805663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,1,power_law_1.2,0.44947199821472167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,8,balanced,0.09385066231091817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,8,balanced,0.09698667128880818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,1,power_law_1.2,1.0894463539123536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,1,power_law_1.2,0.5975679874420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,8,balanced,0.09582933783531189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,1,power_law_1.2,1.3359423637390138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,8,balanced,0.09857066472371419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,1,power_law_1.2,0.6208320140838623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,8,balanced,1.2466613451639812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,8,balanced,0.1204213301340739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,1,power_law_1.2,0.631987190246582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,1,power_law_1.2,1.550598430633545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,8,balanced,0.11916800340016682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,1,power_law_1.2,0.6828608036041259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,1,power_law_1.2,1.9894079208374023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,8,balanced,0.1472053329149882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,1,power_law_1.2,0.6814720153808593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,1,power_law_1.2,2.4337663650512695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,8,balanced,1.5979092915852864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,8,balanced,0.16582933068275452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,1,power_law_1.2,0.7055232048034668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,1,power_law_1.2,3.2942272186279298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,8,balanced,0.23306665817896524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,1,power_law_1.2,0.7474751949310303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,1,power_law_1.2,4.152953720092773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,8,balanced,0.2792533238728841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,1,power_law_1.2,0.7716351985931397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,8,balanced,1.9596586227416992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,1,power_law_1.2,5.03240966796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,1,power_law_1.2,0.7871039867401123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,8,balanced,0.39182400703430176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,1,power_law_1.2,0.8035327911376953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,1,power_law_1.2,7.635327911376953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,8,balanced,0.5007253487904867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,1,power_law_1.2,0.9047167778015137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,1,power_law_1.2,14.658181762695312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.2,0.9661696434020997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,8,balanced,2.951610565185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.2,1.0839167594909669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,8,balanced,0.7146133581797282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.2,1.2035648345947265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.2,1.4653120040893555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.2,1.6894527435302735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,8,balanced,0.9136799971262614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.2,2.1443519592285156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,1,power_law_1.2,2.628447914123535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,1,power_law_1.2,3.5489727020263673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,8,balanced,1.1223093668619792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,8,balanced,5.3136641184488935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,1,power_law_1.2,4.441203308105469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,1,power_law_1.2,5.392140960693359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,1,power_law_1.2,8.121222686767577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,8,balanced,1.7282880147298176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,1,power_law_1.2,15.47039337158203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,8,balanced,3.3049440383911133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,8,balanced,0.09402666489283244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,8,balanced,0.09407466650009155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,8,balanced,0.09132267038027446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,8,balanced,0.09950400392214458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,8,balanced,0.09381332993507385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,8,balanced,0.09082667032877605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,8,balanced,0.09578133622805278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,8,balanced,0.09002133210500081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,8,balanced,0.091839998960495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,8,balanced,0.0883626639842987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,8,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,8,balanced,0.08655466636021932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,8,balanced,0.03422400106986364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,8,balanced,0.08892266949017842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,8,balanced,0.03310399999221166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,8,balanced,0.08370666702588399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,8,balanced,0.0341333324710528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,8,balanced,0.08406399687131245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,8,balanced,0.03472533325354258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,8,balanced,0.08126399914423625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,8,balanced,0.04135466615358988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,8,balanced,0.0687360018491745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,8,balanced,0.08050133287906647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,8,balanced,0.06956266860167186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,8,balanced,0.07606933514277141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,8,balanced,0.07269333302974701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,8,balanced,0.10605866710344951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,8,balanced,0.06869333485762279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,8,balanced,0.09547733267148335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,8,balanced,0.06113600234190623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,8,balanced,0.11873066425323486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,8,balanced,0.10284266869227092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,8,balanced,0.09478400150934856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,8,balanced,0.13271466890970865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,8,balanced,0.07088000078996022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,8,power_law_1.01,0.16299519538879395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,8,balanced,0.10955199599266052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,8,power_law_1.01,0.16173440217971802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,8,balanced,0.08390933275222778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,8,power_law_1.01,0.16176639795303344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,8,balanced,0.09408533573150635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,8,power_law_1.01,0.1600640058517456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,8,balanced,0.13276267051696777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,8,power_law_1.01,0.15931520462036133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,8,balanced,0.15650666753451029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,8,power_law_1.01,0.15550719499588012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,8,balanced,0.22059732675552368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,8,power_law_1.01,0.1466879963874817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,8,power_law_1.01,0.1416383981704712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,8,balanced,0.28036266565322876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,8,power_law_1.01,0.13829120397567748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,8,power_law_1.01,0.10561280250549317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,8,power_law_1.01,0.13533439636230468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,8,balanced,0.39582931995391846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,8,power_law_1.01,0.09832320213317872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,8,power_law_1.01,0.13359999656677246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,8,power_law_1.01,0.09574400186538697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,8,power_law_1.01,0.15249919891357422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,8,power_law_1.01,0.10364799499511719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,8,balanced,0.5039946635564169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,8,power_law_1.01,0.16554880142211914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,8,power_law_1.01,0.10319360494613647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,8,power_law_1.01,0.17552640438079833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,8,power_law_1.01,0.09848319888114929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,8,power_law_1.01,0.19543039798736572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,8,balanced,0.7382720311482748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,8,power_law_1.01,0.09789440035820007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,8,power_law_1.01,0.21447041034698486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,8,power_law_1.01,0.0945855975151062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,8,power_law_1.01,0.246560001373291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,8,power_law_1.01,0.09460480213165283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,8,power_law_1.01,0.3734463930130005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,8,power_law_1.01,0.09033600091934205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,8,power_law_1.01,0.4008639812469482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,8,balanced,0.9643466472625732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,8,power_law_1.01,0.08917120099067688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,8,power_law_1.01,0.6468480110168457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,8,power_law_1.01,0.10632319450378418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,8,power_law_1.01,0.7678783893585205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,8,power_law_1.01,0.11611520051956177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,8,power_law_1.01,1.2270400047302246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,8,power_law_1.01,0.11535359621047973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,8,power_law_1.01,1.549612808227539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,8,balanced,1.4211680094401042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,8,power_law_1.01,0.12362240552902222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,8,power_law_1.01,2.2017343521118162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,8,power_law_1.01,0.12442879676818848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,8,power_law_1.01,2.869766426086426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,8,power_law_1.01,0.14440959692001343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,8,power_law_1.01,0.20205440521240234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,8,power_law_1.01,4.1061054229736325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,8,power_law_1.01,0.2342400074005127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,8,power_law_1.01,4.88598403930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,8,power_law_1.01,0.3173696041107178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,8,balanced,1.864741325378418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,8,power_law_1.01,6.2190399169921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,8,power_law_1.01,0.39013121128082273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,8,power_law_1.01,0.5680511951446533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,8,power_law_1.01,8.970496368408202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,8,power_law_1.01,0.7303103923797607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,8,power_law_1.01,1.0798015594482422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,8,power_law_1.01,17.732147216796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,8,power_law_1.01,1.4216896057128907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,8,balanced,2.3277014096577964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,8,power_law_1.01,2.1184192657470704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,8,power_law_1.01,2.750284767150879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,8,power_law_1.01,3.079360008239746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,8,power_law_1.01,5.247820663452148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,8,power_law_1.01,9.591603088378907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,8,balanced,3.6612319946289062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,8,balanced,7.302330652872722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,8,power_law_1.01,0.09251840114593506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,8,power_law_1.01,0.09542400240898133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,8,power_law_1.01,0.09339519739151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,8,power_law_1.01,0.09132159948348999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,8,power_law_1.01,0.0948032021522522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,8,power_law_1.01,0.08920959830284118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,8,power_law_1.01,0.08871039748191833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,8,power_law_1.01,0.08255360126495362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,8,power_law_1.01,0.030822399258613586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,8,power_law_1.01,0.08145279884338379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,8,power_law_1.01,0.03128319978713989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,8,power_law_1.01,0.08261119723320007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,8,power_law_1.01,0.030611199140548707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,8,power_law_1.01,0.07148799896240235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,8,power_law_1.01,0.03152639865875244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,8,power_law_1.01,0.08756480216979981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,8,power_law_1.01,0.03237119913101196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,8,power_law_1.01,0.09818239808082581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,8,power_law_1.01,0.03857280015945434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,8,power_law_1.01,0.09726719856262207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,8,power_law_1.01,0.06723200082778931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,8,power_law_1.01,0.11075199842453003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,8,power_law_1.01,0.06659839749336242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,8,power_law_1.01,0.10257920026779174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,8,balanced,0.17596266667048135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,8,power_law_1.01,0.07378559708595275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,8,power_law_1.01,0.1356927990913391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,8,power_law_1.01,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,8,power_law_1.01,0.1599679946899414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,8,balanced,0.21608533461888632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,8,power_law_1.01,0.05905280113220215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.01,0.18556159734725952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,8,power_law_1.01,0.09850879907608032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.01,0.26030080318450927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,8,power_law_1.01,0.09099519848823548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.01,0.3300800085067749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,8,balanced,0.2942506670951843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,8,power_law_1.01,0.0679040014743805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.01,0.4787456035614014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,8,power_law_1.01,0.10780160427093506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.01,0.5750592231750489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,8,power_law_1.2,0.16151039600372313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,8,balanced,0.36834665139516193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,8,power_law_1.01,0.08279039859771728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.01,0.7758207798004151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,8,power_law_1.2,0.1652672052383423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,8,power_law_1.01,0.09216639995574952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.01,1.1938048362731934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,8,power_law_1.2,0.15949440002441406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,8,power_law_1.01,0.1313279986381531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,8,power_law_1.2,0.1664512038230896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.01,1.609574317932129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.01,0.1554111957550049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,8,power_law_1.2,0.16027519702911378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.01,2.040518379211426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,8,balanced,0.5354559818903605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.01,0.21829121112823485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,8,power_law_1.2,0.15063680410385133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,8,power_law_1.01,2.5775680541992188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.01,0.279200005531311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,8,power_law_1.2,0.1500480055809021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,8,power_law_1.01,4.577836990356445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,8,power_law_1.2,0.14217599630355834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.01,0.3964416027069092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,8,power_law_1.2,0.14030079841613768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.01,0.5041920185089112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,8,power_law_1.01,8.7500732421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,8,balanced,0.696021318435669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,8,power_law_1.2,0.13800959587097167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.01,0.7406847953796387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,8,power_law_1.2,0.12948479652404785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.01,0.9633343696594239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,8,power_law_1.2,0.17475839853286743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.01,1.4138367652893067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,8,power_law_1.2,0.16682239770889282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.01,1.8660480499267578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,8,power_law_1.2,0.1637312054634094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,8,power_law_1.01,2.325331115722656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,8,power_law_1.2,0.19852160215377807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,8,balanced,0.8512372970581055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,8,power_law_1.2,0.21189119815826415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,8,power_law_1.01,3.6623809814453123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,8,power_law_1.2,0.2556159973144531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,8,power_law_1.01,7.277254486083985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,8,power_law_1.2,0.36321918964385985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,8,power_law_1.2,0.41719679832458495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,8,power_law_1.2,0.6467904090881348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,8,power_law_1.2,0.8380928039550781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,8,power_law_1.2,1.1703807830810546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,8,balanced,1.3340320587158203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,8,power_law_1.2,1.5286335945129395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,8,power_law_1.2,2.009267234802246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,8,power_law_1.2,2.8008256912231446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,8,power_law_1.2,3.517241668701172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,8,power_law_1.2,5.091897583007812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,8,power_law_1.2,5.724863815307617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,8,power_law_1.2,9.513132476806641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,8,power_law_1.2,17.85017547607422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,8,balanced,2.6140480041503906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,8,power_law_1.2,0.10657279491424561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,8,power_law_1.2,0.09687039852142335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,8,power_law_1.2,0.1004480004310608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,8,power_law_1.2,0.10094720125198364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,8,power_law_1.2,0.0997439980506897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,8,power_law_1.2,0.09802240133285522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,8,power_law_1.2,0.09479039907455444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,8,power_law_1.2,0.09423360228538513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,8,power_law_1.2,0.0893887996673584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,8,power_law_1.2,0.09164800047874451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,8,power_law_1.2,0.08773120045661927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,8,power_law_1.2,0.08952320218086243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,8,power_law_1.2,0.08600959777832032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,8,power_law_1.2,0.08726400136947632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,8,power_law_1.2,0.08865919709205627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,8,power_law_1.2,0.11393920183181763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,8,power_law_1.2,0.09123839735984803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,8,power_law_1.2,0.11566720008850098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,8,power_law_1.2,0.09086719751358033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,8,power_law_1.2,0.11111040115356445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,8,power_law_1.2,0.08784639835357666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,8,power_law_1.2,0.13568639755249023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,8,power_law_1.2,0.08453119993209839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,8,power_law_1.2,0.13640320301055908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,8,power_law_1.2,0.08119040131568908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,8,power_law_1.2,0.14742399454116822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,8,power_law_1.2,0.07602559924125671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,8,power_law_1.2,0.20278398990631102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,8,power_law_1.2,0.06863359808921814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,8,power_law_1.2,0.23790719509124755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,8,power_law_1.2,0.08960639834403991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,8,power_law_1.2,0.314246392250061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,8,power_law_1.2,0.08916479945182801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,8,power_law_1.2,0.42808961868286133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,8,power_law_1.2,0.09060479998588562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,8,power_law_1.2,0.537830400466919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,8,power_law_1.2,0.11923199892044067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,8,power_law_1.2,0.6903744220733643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,8,power_law_1.2,0.03280639946460724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,8,power_law_1.2,0.11723519563674926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,8,power_law_1.2,1.0841535568237304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,8,power_law_1.2,0.032953599095344545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,8,power_law_1.2,0.12551679611206054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,8,power_law_1.2,1.4084223747253417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,8,power_law_1.2,0.03158400058746338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,8,power_law_1.2,0.17226879596710204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,8,power_law_1.2,1.9633472442626954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,8,power_law_1.2,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.2,0.19160959720611573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,8,power_law_1.2,2.7811967849731447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,8,power_law_1.2,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.2,0.252512001991272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,8,power_law_1.2,0.03743360042572021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,8,power_law_1.2,3.4203006744384767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.2,0.3235775947570801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,8,power_law_1.2,0.06778240203857422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,4,balanced,0.09872000416119893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.2,0.4752448081970215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,8,power_law_1.2,5.303526306152344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,8,power_law_1.2,0.06893439888954163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,4,balanced,0.10240000486373901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.2,0.6176703929901123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,8,power_law_1.2,0.07192959785461425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,4,balanced,0.13507200280825296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,8,power_law_1.2,10.31934051513672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.2,0.8735103607177734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,4,balanced,0.13774933417638144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,8,power_law_1.2,0.06826239824295044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.2,1.1452287673950194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,4,balanced,0.14069333672523499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,8,power_law_1.2,0.059699201583862306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.2,1.765228843688965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,4,balanced,0.13369066516558328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,8,power_law_1.2,0.10346879959106445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,4,balanced,0.14355732997258505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.2,2.2108287811279297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,8,power_law_1.2,0.09409279823303222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,4,balanced,0.1313706636428833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,8,power_law_1.2,0.06932479739189149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,8,power_law_1.2,2.597158432006836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,4,balanced,0.1336426635583242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,8,power_law_1.2,0.10823680162429809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,8,power_law_1.2,4.5803264617919925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,4,balanced,0.13891200224558511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,8,power_law_1.2,0.0812287986278534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,4,balanced,0.12875200311342874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,8,power_law_1.2,0.09152640104293823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,8,power_law_1.2,8.264320373535156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,4,balanced,0.12100266416867574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,8,power_law_1.2,0.13073279857635497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,4,balanced,0.12293333808581035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.2,0.1567744016647339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,4,balanced,0.11834667126337688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.2,0.21965439319610597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,4,balanced,0.11529067158699036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.2,0.2796799898147583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,4,balanced,0.11219732960065205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.2,0.3964672088623047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,4,balanced,0.1144480009873708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.2,0.5024320125579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,4,balanced,0.15985600153605142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.2,0.7344319820404053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,4,balanced,0.14478933811187744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.2,0.9627200126647949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,4,balanced,0.2042293349901835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.2,1.4161727905273438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.2,1.8596736907958984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,4,balanced,0.2300586700439453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,8,power_law_1.2,2.3138496398925783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,4,balanced,0.32338666915893555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,8,power_law_1.2,3.6759616851806642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,4,balanced,0.3999840021133423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,8,power_law_1.2,7.296211242675781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,4,balanced,0.5554346640904745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,4,balanced,0.6999626954396566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,4,balanced,0.986512025197347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,4,balanced,1.2696479956309001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,4,balanced,1.5539466540018718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,4,balanced,2.396517276763916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,4,balanced,4.620688120524089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,4,balanced,0.06656000018119812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,4,balanced,0.06941866874694824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,4,balanced,0.09485333164532979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,4,balanced,0.08983467022577922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,4,balanced,0.0867199997107188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,4,balanced,0.09204799930254619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,4,balanced,0.09542933106422424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,4,balanced,0.09319999814033508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,4,balanced,0.0647680014371872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,4,balanced,0.08760533730189006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,4,balanced,0.06213866670926412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,4,balanced,0.08659733335177104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,4,balanced,0.07969599962234497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,4,balanced,0.09028266867001851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,4,balanced,0.0800799975792567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,4,balanced,0.0844693382581075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,4,balanced,0.08402132987976074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,4,balanced,0.08783466617266338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,4,balanced,0.07718400160471599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,4,balanced,0.08244266609350841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,4,balanced,0.08180800080299377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,4,balanced,0.08514666557312012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,4,balanced,0.08036266764005025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,4,balanced,0.08621866504351298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,4,balanced,0.08161599934101105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,4,balanced,0.08772266904513042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,4,balanced,0.08363733688990276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,4,balanced,0.11301333705584209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,4,balanced,0.07553066809972127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,4,balanced,0.10338667035102844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,4,balanced,0.0779306689898173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,4,balanced,0.0787360022465388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,4,balanced,0.13992533087730408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,4,balanced,0.07220800220966339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,4,balanced,0.1567093332608541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,4,balanced,0.08045333127180736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,4,balanced,0.07326399783293407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,4,balanced,0.2087093393007914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,4,balanced,0.07288533449172974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,4,balanced,0.2581706643104553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,4,balanced,0.10199999809265137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,4,balanced,0.024101334313551586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,4,balanced,0.0934826632340749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,4,balanced,0.02568000058333079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,4,balanced,0.3670399983723958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,4,balanced,0.02436800052722295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,4,balanced,0.11728533109029134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,4,balanced,0.0266239990790685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,4,balanced,0.13689600427945456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,4,balanced,0.026869334280490875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,4,balanced,0.46490665276845294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,4,balanced,0.028245332340399425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,4,balanced,0.17811199029286703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,4,balanced,0.03682666768630346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,4,balanced,0.031221332649389904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,4,balanced,0.06107200185457865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,4,balanced,0.667914628982544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,4,balanced,0.21523199478785196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,4,balanced,0.06131199995676676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,4,balanced,0.06108266611893972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,4,balanced,0.2995786666870117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,4,balanced,0.06341866652170818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,4,balanced,0.06265600025653839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,4,balanced,0.8708266417185465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,4,balanced,0.04850133260091146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,4,balanced,0.37941332658131915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,4,balanced,0.06419200201829274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,4,balanced,0.0576800008614858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,4,balanced,0.05919999877611796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,4,balanced,0.5422133207321167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,4,balanced,1.0755733648935955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,4,balanced,0.08379733562469482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,4,balanced,0.0944106678167979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,4,balanced,0.13060266772905985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,4,balanced,0.7022506395975748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,4,balanced,0.1616426706314087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,4,balanced,0.227567990620931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,4,balanced,1.6809546152750652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,4,balanced,0.2862773338953654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,4,balanced,0.8731520175933838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,4,balanced,0.4137279987335205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,4,balanced,0.5379999876022339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,4,balanced,1.3687200546264648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,4,balanced,0.7843519846598307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,4,balanced,3.2990134557088218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,4,balanced,1.031610647837321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,4,power_law_1.01,0.12162560224533081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,4,balanced,2.704122543334961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,4,power_law_1.01,0.11868159770965576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,4,balanced,1.279525359471639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,4,power_law_1.01,0.12265599966049194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,4,power_law_1.01,0.14213119745254515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,4,power_law_1.01,0.13734400272369385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,4,power_law_1.01,0.1336192011833191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,4,power_law_1.01,0.13236479759216307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,4,balanced,2.0133919715881348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,4,power_law_1.01,0.1335487961769104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,4,power_law_1.01,0.1275904059410095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,4,power_law_1.01,0.12309119701385499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,4,power_law_1.01,0.11845760345458985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,4,power_law_1.01,0.13414399623870848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,4,power_law_1.01,0.13644800186157227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,4,power_law_1.01,0.1442944049835205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,4,power_law_1.01,0.16200319528579712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,4,power_law_1.01,0.1664255976676941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,4,power_law_1.01,0.22287359237670898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,4,power_law_1.01,0.23419520854949952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,4,power_law_1.01,0.2942271947860718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,4,power_law_1.01,0.3853503942489624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,4,balanced,3.989269256591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,4,power_law_1.01,0.46204161643981934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,4,power_law_1.01,0.7385024070739746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,4,power_law_1.01,0.7338751792907715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,4,power_law_1.01,1.1030464172363281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,4,power_law_1.01,1.6253568649291992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,4,power_law_1.01,1.981100845336914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,4,power_law_1.01,3.1112768173217775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,4,power_law_1.01,4.008556747436524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,4,power_law_1.01,5.6735679626464846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,4,power_law_1.01,13.838067626953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,4,power_law_1.01,0.07289599776268005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,4,power_law_1.01,0.07912319898605347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,4,power_law_1.01,0.08296959996223449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,4,power_law_1.01,0.0870464026927948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,4,power_law_1.01,0.08999680280685425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,4,power_law_1.01,0.08380159735679626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,4,power_law_1.01,0.08761600255966187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,4,power_law_1.01,0.08529919981956482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,4,power_law_1.01,0.07564160227775574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,4,power_law_1.01,0.0852735996246338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,4,power_law_1.01,0.0772607982158661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,4,power_law_1.01,0.08472959995269776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,4,power_law_1.01,0.07364479899406433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,4,power_law_1.01,0.08256000280380249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,4,power_law_1.01,0.0781503975391388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,4,power_law_1.01,0.08376320004463196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,4,power_law_1.01,0.0804032027721405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,4,power_law_1.01,0.08906880021095276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,4,power_law_1.01,0.07955200076103211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,4,power_law_1.01,0.09393919706344604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,4,power_law_1.01,0.08191999793052673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,4,power_law_1.01,0.10840320587158203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,4,power_law_1.01,0.08075519800186157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,4,power_law_1.01,0.10026880502700805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,4,power_law_1.01,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,4,power_law_1.01,0.13528319597244262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,4,power_law_1.01,0.07543039917945862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,4,power_law_1.01,0.07505919933319091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,4,power_law_1.01,0.14592000246047973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,4,power_law_1.01,0.09148160219192505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,4,power_law_1.01,0.17519999742507936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,4,power_law_1.01,0.08751360177993775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,4,power_law_1.01,0.20904319286346434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,4,power_law_1.01,0.08851199746131896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,4,power_law_1.01,0.2729408025741577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,4,power_law_1.01,0.09994239807128906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,4,power_law_1.01,0.3690176010131836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,4,power_law_1.01,0.09840000271797181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,4,power_law_1.01,0.5407423973083496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,4,power_law_1.01,0.1115839958190918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,4,power_law_1.01,0.6964159965515136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,4,power_law_1.01,0.1382848024368286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,4,power_law_1.01,0.7890751838684082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.01,0.1533504009246826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,4,power_law_1.01,1.6173248291015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.01,0.19174400568008423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,4,power_law_1.01,1.4989184379577636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.01,0.23878400325775145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,4,power_law_1.01,2.316646385192871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.01,0.3630016088485718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,4,power_law_1.01,2.8710208892822267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,4,power_law_1.01,0.02353920042514801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.01,0.40904960632324217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,4,power_law_1.01,0.023212799429893495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.01,0.665337610244751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,4,power_law_1.01,7.1684928894042965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,4,power_law_1.01,0.022745600342750548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.01,0.9370880126953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,4,power_law_1.01,0.025920000672340394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.01,1.3870271682739257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,4,power_law_1.01,0.02595199942588806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.01,1.6473407745361328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,4,power_law_1.01,0.02693760097026825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,4,power_law_1.01,2.2322816848754883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,4,power_law_1.01,0.038771200180053714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,4,power_law_1.01,3.830547332763672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,4,power_law_1.01,0.029158401489257812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,4,power_law_1.01,0.05975040197372437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,4,power_law_1.01,6.91817626953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,4,power_law_1.01,0.05939199924468994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,4,power_law_1.01,0.05994880199432373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,4,power_law_1.01,0.06510720252990723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,4,power_law_1.01,0.059648001194000246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,4,power_law_1.01,0.04689919948577881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,4,power_law_1.01,0.06509439945220948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,4,power_law_1.01,0.056595200300216676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,4,power_law_1.01,0.058899199962615965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,4,power_law_1.2,0.12117120027542114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,4,power_law_1.01,0.0828927993774414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,4,power_law_1.2,0.13162239789962768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.01,0.09345920085906982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,4,power_law_1.2,0.12255359888076782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.01,0.13052159547805786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,4,power_law_1.2,0.1343999981880188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.01,0.16299519538879395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,4,power_law_1.2,0.13895679712295533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.01,0.2248447895050049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,4,power_law_1.2,0.14058879613876343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.01,0.28657920360565187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,4,power_law_1.2,0.1292672038078308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.01,0.41179518699645995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,4,power_law_1.2,0.13005440235137938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.01,0.5371967792510987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,4,power_law_1.2,0.12662400007247926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.01,0.7832511901855469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,4,power_law_1.2,0.12462719678878784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.01,1.0285951614379882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,4,power_law_1.2,0.12006399631500245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,4,power_law_1.01,1.2786304473876953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,4,power_law_1.2,0.12233599424362182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,4,power_law_1.01,2.004761505126953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,4,power_law_1.2,0.137881600856781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,4,power_law_1.01,3.9911678314208983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,4,power_law_1.2,0.15296000242233276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,4,power_law_1.2,0.16989439725875854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,4,power_law_1.2,0.15807360410690308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,4,power_law_1.2,0.18488320112228393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,4,power_law_1.2,0.26086399555206297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,4,power_law_1.2,0.30570240020751954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,4,power_law_1.2,0.43116159439086915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,4,power_law_1.2,0.5920447826385498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,4,power_law_1.2,0.6934463977813721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,4,power_law_1.2,0.9310208320617676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,4,power_law_1.2,1.4723008155822754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,4,power_law_1.2,1.8073087692260743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,4,power_law_1.2,2.811795234680176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,4,power_law_1.2,2.960383987426758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,4,power_law_1.2,4.157113647460937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,4,power_law_1.2,6.898515319824218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,4,power_law_1.2,12.219987487792968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,4,power_law_1.2,0.074726402759552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,4,power_law_1.2,0.08124799728393554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,4,power_law_1.2,0.07768959999084472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,4,power_law_1.2,0.08543999791145325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,4,power_law_1.2,0.0832319974899292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,4,power_law_1.2,0.08980479836463928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,4,power_law_1.2,0.07483519911766053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,4,power_law_1.2,0.08988800048828124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,4,power_law_1.2,0.08566399812698364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,4,power_law_1.2,0.0775168001651764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,4,power_law_1.2,0.08503040075302123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,4,power_law_1.2,0.07470080256462097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,4,power_law_1.2,0.08401280045509338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,4,power_law_1.2,0.08153600096702576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,4,power_law_1.2,0.08243839740753174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,4,power_law_1.2,0.08051199913024902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,4,power_law_1.2,0.08673279881477355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,4,power_law_1.2,0.07923200130462646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,4,power_law_1.2,0.09490560293197632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,4,power_law_1.2,0.08069760203361512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,4,power_law_1.2,0.09174399971961975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,4,power_law_1.2,0.08117120265960694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,4,power_law_1.2,0.11946879625320435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,4,power_law_1.2,0.07474560141563416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,4,power_law_1.2,0.11420799493789673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,4,power_law_1.2,0.07548159956932068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,4,power_law_1.2,0.13358080387115479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,4,power_law_1.2,0.07559040188789368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,4,power_law_1.2,0.15240319967269897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,4,power_law_1.2,0.08681600093841553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,4,power_law_1.2,0.20639359951019287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,4,power_law_1.2,0.08368639945983887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,4,power_law_1.2,0.23283839225769043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,4,power_law_1.2,0.08768640160560608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,4,power_law_1.2,0.2910847902297974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,4,power_law_1.2,0.10463999509811402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,4,power_law_1.2,0.47032961845397947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,4,power_law_1.2,0.10206719636917114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,4,power_law_1.2,0.5407104015350341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,4,power_law_1.2,0.11909120082855225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,4,power_law_1.2,0.8224575996398926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,4,power_law_1.2,0.13848320245742798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,4,power_law_1.2,0.8858431816101074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.2,0.14844800233840943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,4,power_law_1.2,1.609119987487793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.2,0.2035072088241577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.2,0.2354048013687134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,4,power_law_1.2,2.0012544631958007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.2,0.3864192008972168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,4,power_law_1.2,2.2385663986206055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.2,0.442412805557251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,4,power_law_1.2,0.02207999974489212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,4,power_law_1.2,3.775302505493164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.2,0.7245696067810059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,4,power_law_1.2,0.023340800404548646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,4,power_law_1.2,8.327616119384766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,4,power_law_1.2,0.02258560061454773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.2,0.9067456245422363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,4,power_law_1.2,0.02620159983634949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.2,1.3649984359741212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,4,power_law_1.2,0.026399999856948853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.2,1.7198144912719726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,4,power_law_1.2,0.027718400955200194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,4,power_law_1.2,2.497395133972168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,4,power_law_1.2,0.037625598907470706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,4,power_law_1.2,4.410771179199219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,4,power_law_1.2,0.030419200658798218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,4,power_law_1.2,0.060166400671005246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,4,power_law_1.2,8.050656127929688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,4,power_law_1.2,0.060102397203445436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,4,power_law_1.2,0.059622400999069215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,4,power_law_1.2,0.06272000074386597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,4,power_law_1.2,0.06299520134925843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,4,power_law_1.2,0.04624640047550201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,4,power_law_1.2,0.061427199840545656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,4,power_law_1.2,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,4,power_law_1.2,0.058534401655197146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,4,power_law_1.2,0.08300799727439881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.2,0.09218559861183166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.2,0.1297727942466736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.2,0.16095999479293824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.2,0.22580480575561523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.2,0.28740479946136477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,2,balanced,0.06999999781449635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,2,balanced,0.0823520024617513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.2,0.41199359893798826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,2,balanced,0.11540266871452332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.2,0.5387904167175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,2,balanced,0.11359467109044392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.2,0.7817535877227784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,2,balanced,0.11202133695284526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.2,1.0291007995605468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,2,balanced,0.11829333504041036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,4,power_law_1.2,1.2730688095092773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,2,balanced,0.11563733220100403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,2,balanced,0.1127786636352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,4,power_law_1.2,2.0077823638916015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,2,balanced,0.12220799922943115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,4,power_law_1.2,3.9907199859619142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,2,balanced,0.11517332990964253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,2,balanced,0.10873599847157796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,2,balanced,0.11080533266067505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,2,balanced,0.10705600182215373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,2,balanced,0.10156266887982686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,2,balanced,0.10735467076301575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,2,balanced,0.10666132966677348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,2,balanced,0.10790933171908061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,2,balanced,0.15541332960128784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,2,balanced,0.14139200250307718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,2,balanced,0.1830400029818217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,2,balanced,0.21420266230901083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,2,balanced,0.28545600175857544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,2,balanced,0.34730132420857746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,2,balanced,0.506223996480306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,2,balanced,0.6452320019404093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,2,balanced,0.9388319651285807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,2,balanced,1.2396213213602703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,2,balanced,1.5330079396565754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,2,balanced,0.04839999973773956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,2,balanced,0.055717334151268005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,2,balanced,0.08726400136947632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,2,balanced,0.07834666470686595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,2,balanced,0.07387199997901917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,2,balanced,2.432117303212484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,2,balanced,0.07519466678301494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,2,balanced,0.08851733803749084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,2,balanced,0.049072002371152244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,2,balanced,0.0816480020682017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,2,balanced,0.055589333176612854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,2,balanced,0.08572799960772197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,2,balanced,0.07737599809964497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,2,balanced,0.08176533381144206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,2,balanced,0.07605333129564922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,2,balanced,0.07969599962234497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,2,balanced,0.07705066601435344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,2,balanced,0.07432533303896587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,2,balanced,0.07724800209204356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,2,balanced,0.08156799773375194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,2,balanced,0.07713066538174947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,2,balanced,0.07817066709200542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,2,balanced,0.07569066683451335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,2,balanced,0.07998399933179219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,2,balanced,0.08207466701666515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,2,balanced,0.08168533444404602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,2,balanced,0.08193066716194153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,2,balanced,0.07628266513347626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,2,balanced,0.08473599950472514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,2,balanced,4.824458758036296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,2,balanced,0.08429333567619324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,2,balanced,0.1083733340104421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,2,balanced,0.07402133444945018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,2,balanced,0.10829866925875346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,2,balanced,0.07655466596285503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,2,balanced,0.13478933771451315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,2,balanced,0.07517333328723907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,2,balanced,0.07390933235486348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,2,balanced,0.15845866998036703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,2,balanced,0.07266666491826375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,2,balanced,0.2137653430302938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,2,balanced,0.10893332958221436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,2,balanced,0.10165866216023763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,2,balanced,0.25982399781545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,2,balanced,0.1213653286298116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,2,balanced,0.14320533474286398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,2,balanced,0.36657599608103436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,2,balanced,0.021520001192887623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,2,balanced,0.024245334168275196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,2,balanced,0.18826133012771606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,2,balanced,0.024671999116738636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,2,balanced,0.4675946633021037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,2,balanced,0.024058667321999867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,2,balanced,0.23471999168395996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,2,balanced,0.026906666656335194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,2,balanced,0.02736533433198929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,2,balanced,0.028560000161329906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,2,balanced,0.32262933254241943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,2,balanced,0.6756213506062826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,2,balanced,0.028543998797734577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,2,balanced,0.030762667457262676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,2,balanced,0.028768000503381092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,2,balanced,0.40760000546773273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,2,balanced,0.03090133269627889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,2,balanced,0.8909546534220377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,2,balanced,0.05754133562246958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,2,balanced,0.05752533177534739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,2,balanced,0.0598880002895991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,2,balanced,0.5906613270441691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,2,balanced,0.0470773329337438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,2,balanced,0.04509866734345754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,2,balanced,1.1065173149108887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,2,balanced,0.046538665890693665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,2,balanced,0.058431997895240784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,2,balanced,0.7720106442769369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,2,balanced,0.06176533301671346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,2,balanced,0.08480000495910645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,2,balanced,0.09948800007502238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,2,balanced,1.7453546524047852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,2,balanced,0.1376053293546041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,2,balanced,0.9478399753570557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,2,balanced,0.17082132895787558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,2,balanced,0.24381866057713827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,2,balanced,1.4855999946594238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,2,balanced,0.3142613371213277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,2,balanced,0.4531466563542684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,2,balanced,3.4150772094726562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,2,balanced,0.5930933157602946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,2,balanced,2.9314400355021157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,2,balanced,0.735205332438151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,2,balanced,1.1549866994222004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,2,power_law_1.01,0.07303680181503296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,2,power_law_1.01,0.07440639734268188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,2,power_law_1.01,0.09292160272598267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,2,power_law_1.01,0.11278079748153687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,2,balanced,2.2730239232381186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,2,power_law_1.01,0.11386239528656006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,2,power_law_1.01,0.12245759963989258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,2,power_law_1.01,0.1156607985496521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,2,power_law_1.01,0.1186560034751892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,2,power_law_1.01,0.11939200162887573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,2,power_law_1.01,0.11496959924697876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,2,power_law_1.01,0.11579519510269165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,2,power_law_1.01,0.1192512035369873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,2,power_law_1.01,0.12106879949569702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,2,power_law_1.01,0.12611839771270753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,2,power_law_1.01,0.14362879991531372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,2,power_law_1.01,0.1479423999786377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,2,power_law_1.01,0.1513856053352356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,2,power_law_1.01,0.19555200338363649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,2,power_law_1.01,0.21389439105987548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,2,power_law_1.01,0.26005120277404786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,2,power_law_1.01,0.31021440029144287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,2,power_law_1.01,0.42803201675415037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,2,power_law_1.01,0.5945087909698487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,2,power_law_1.01,0.8349823951721191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,2,power_law_1.01,1.0374655723571777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,2,power_law_1.01,1.5536959648132325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,2,power_law_1.01,1.9327936172485352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,2,power_law_1.01,2.224857521057129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,2,power_law_1.01,3.9030078887939452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,2,power_law_1.01,7.363565063476562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,2,power_law_1.01,0.05911039710044861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,2,power_law_1.01,0.05922560095787048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,2,power_law_1.01,0.05817599892616272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,2,power_law_1.01,0.08204159736633301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,2,power_law_1.01,0.07875199913978577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,2,power_law_1.01,0.07664639949798584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,2,power_law_1.01,0.08135039806365967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,2,power_law_1.01,0.07998719811439514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,2,power_law_1.01,0.07787520289421082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,2,power_law_1.01,0.0793727993965149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,2,power_law_1.01,0.07780479788780212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,2,power_law_1.01,0.05616000294685364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,2,power_law_1.01,0.08429440259933471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,2,power_law_1.01,0.05854079723358154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,2,power_law_1.01,0.08945919871330262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,2,power_law_1.01,0.06288639903068542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,2,power_law_1.01,0.0898751974105835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,2,power_law_1.01,0.0802944004535675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,2,power_law_1.01,0.10092159509658813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,2,power_law_1.01,0.07861760258674622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,2,power_law_1.01,0.10368640422821045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,2,power_law_1.01,0.07589120268821717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,2,power_law_1.01,0.11207679510116578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,2,power_law_1.01,0.07753599882125854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,2,power_law_1.01,0.1263808012008667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,2,power_law_1.01,0.07600640058517456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,2,power_law_1.01,0.14419200420379638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,2,power_law_1.01,0.0742143988609314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,2,power_law_1.01,0.18699519634246825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,2,power_law_1.01,0.0753216028213501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,2,power_law_1.01,0.2028223991394043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,2,power_law_1.01,0.0748095989227295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,2,power_law_1.01,0.07843199968338013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,2,power_law_1.01,0.2797823905944824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,2,power_law_1.01,0.08166400194168091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,2,power_law_1.01,0.3592128038406372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,2,power_law_1.01,0.08387200236320495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,2,power_law_1.01,0.525932788848877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,2,power_law_1.01,0.09151999950408936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,2,power_law_1.01,0.7183167934417725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,2,power_law_1.01,0.09899520277976989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,2,power_law_1.01,1.0609408378601075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,2,power_law_1.01,0.10812159776687622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,2,power_law_1.01,1.329036808013916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,2,power_law_1.01,0.12009600400924683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,2,power_law_1.01,1.6528383255004884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.01,0.13515520095825195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,2,power_law_1.01,2.4725759506225584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.01,0.1684224009513855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,2,power_law_1.01,4.815571212768555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.01,0.198361599445343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.01,0.25430400371551515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,2,power_law_1.01,0.021184000372886657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.01,0.32071681022644044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,2,power_law_1.01,0.02260479927062988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.01,0.4770751953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,2,power_law_1.01,0.02237440049648285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.01,0.5559296131134033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,2,power_law_1.01,0.02186879962682724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.01,0.8052672386169434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,2,power_law_1.01,0.025971201062202454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.01,1.2985471725463866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,2,power_law_1.01,0.027116799354553224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,2,power_law_1.01,0.026476800441741943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,2,power_law_1.01,1.5186816215515138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,2,power_law_1.01,0.02685439884662628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,2,power_law_1.01,2.3730239868164062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,2,power_law_1.01,0.029190400242805482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,2,power_law_1.01,4.4320320129394535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,2,power_law_1.01,0.02876160144805908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,2,power_law_1.01,0.02922239899635315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,2,power_law_1.01,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,2,power_law_1.01,0.056454402208328244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,2,power_law_1.01,0.05882880091667175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,2,power_law_1.01,0.04599039852619171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,2,power_law_1.01,0.04434559941291809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,2,power_law_1.01,0.0455487996339798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,2,power_law_1.01,0.05727360248565674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.01,0.06047999858856201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.01,0.08528000116348267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.01,0.09893760085105896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.01,0.13602559566497802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.01,0.17015039920806885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.01,0.24106240272521973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,2,power_law_1.2,0.0720575988292694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.01,0.3127552032470703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,2,power_law_1.2,0.08648319840431214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,2,power_law_1.2,0.09509119987487794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.01,0.45157761573791505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,2,power_law_1.2,0.11635839939117432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.01,0.5936831951141357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,2,power_law_1.2,0.1210368037223816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,2,power_law_1.01,0.7357888221740723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,2,power_law_1.2,0.11804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,2,power_law_1.01,1.1497599601745605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,2,power_law_1.2,0.11929600238800049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,2,power_law_1.01,2.2694015502929688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,2,power_law_1.2,0.11485439538955688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,2,power_law_1.2,0.11880320310592651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,2,power_law_1.2,0.11463040113449097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,2,power_law_1.2,0.11267839670181275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,2,power_law_1.2,0.12988799810409546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,2,power_law_1.2,0.1293951988220215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,2,power_law_1.2,0.13074560165405275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,2,power_law_1.2,0.14423680305480957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,2,power_law_1.2,0.15140479803085327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,2,power_law_1.2,0.16228480339050294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,2,power_law_1.2,0.18568320274353028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,2,power_law_1.2,0.22014079093933106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,2,power_law_1.2,0.26806399822235105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,2,power_law_1.2,0.2898240089416504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,2,power_law_1.2,0.4098495960235596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,2,power_law_1.2,0.577888011932373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,2,power_law_1.2,0.7805503845214844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,2,power_law_1.2,1.0180480003356933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,2,power_law_1.2,1.6993215560913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,2,power_law_1.2,2.011180877685547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,2,power_law_1.2,2.7428415298461912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,2,power_law_1.2,3.7870014190673826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,2,power_law_1.2,7.382150268554687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,2,power_law_1.2,0.060838401317596436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,2,power_law_1.2,0.05192959904670715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,2,power_law_1.2,0.06235520243644714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,2,power_law_1.2,0.07738879919052125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,2,power_law_1.2,0.0749504029750824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,2,power_law_1.2,0.0809664011001587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,2,power_law_1.2,0.07976959943771363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,2,power_law_1.2,0.07934719920158387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,2,power_law_1.2,0.07975680232048035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,2,power_law_1.2,0.05297279953956604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,2,power_law_1.2,0.07875199913978577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,2,power_law_1.2,0.053504002094268796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,2,power_law_1.2,0.07854080200195312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,2,power_law_1.2,0.0831488013267517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,2,power_law_1.2,0.06034560203552246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,2,power_law_1.2,0.08711040019989014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,2,power_law_1.2,0.07438719868659974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,2,power_law_1.2,0.08847360014915466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,2,power_law_1.2,0.07667199969291687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,2,power_law_1.2,0.09690880179405212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,2,power_law_1.2,0.07949439883232116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,2,power_law_1.2,0.10706559419631959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,2,power_law_1.2,0.07382400035858154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,2,power_law_1.2,0.10651520490646363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,2,power_law_1.2,0.07807999849319458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,2,power_law_1.2,0.07423999905586243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,2,power_law_1.2,0.13150719404220582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,2,power_law_1.2,0.07712640166282654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,2,power_law_1.2,0.15334399938583373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,2,power_law_1.2,0.07657600045204163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,2,power_law_1.2,0.19414399862289428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,2,power_law_1.2,0.07761279940605163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,2,power_law_1.2,0.22458879947662352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,2,power_law_1.2,0.08388479948043823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,2,power_law_1.2,0.2799743890762329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,2,power_law_1.2,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,2,power_law_1.2,0.3859008073806763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,2,power_law_1.2,0.09404799938201905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,2,power_law_1.2,0.5500415802001953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,2,power_law_1.2,0.09742720127105713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,2,power_law_1.2,0.7289279937744141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,2,power_law_1.2,0.10055040121078491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,2,power_law_1.2,0.8867072105407715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,2,power_law_1.2,0.12097280025482178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,2,power_law_1.2,1.2788607597351074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.2,0.1366528034210205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,2,power_law_1.2,1.5786815643310548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.2,0.1784767985343933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,2,power_law_1.2,2.6338687896728517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,2,power_law_1.2,0.02170879989862442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.2,0.21729280948638915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,2,power_law_1.2,0.02250880002975464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,2,power_law_1.2,4.6827392578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.2,0.2869375944137573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,2,power_law_1.2,0.022617599368095397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.2,0.3598207950592041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,2,power_law_1.2,0.02207999974489212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.2,0.4729919910430908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,2,power_law_1.2,0.023897600173950196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.2,0.583129596710205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,2,power_law_1.2,0.026822400093078614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.2,0.9055616378784179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,2,power_law_1.2,0.026931199431419372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.2,1.1104767799377442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,2,power_law_1.2,0.027136000990867614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,2,power_law_1.2,1.286457633972168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,2,power_law_1.2,0.02940160036087036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,2,power_law_1.2,0.027513599395751952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,2,power_law_1.2,2.117145538330078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,2,power_law_1.2,0.02974080145359039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,2,power_law_1.2,5.045292663574219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,2,power_law_1.2,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,2,power_law_1.2,0.05802239775657654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,2,power_law_1.2,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,2,power_law_1.2,0.04572800099849701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,2,power_law_1.2,0.04438399970531463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,2,power_law_1.2,0.04521600008010864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,2,power_law_1.2,0.05851519703865051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.2,0.060262399911880496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.2,0.08499839901924133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.2,0.10027519464492798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.2,0.13782399892807007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.2,0.1682752013206482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.2,0.24137599468231202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.2,0.3147775888442993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.2,0.45241599082946776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,1,balanced,0.05574400226275126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.2,0.5961343765258789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,1,balanced,0.07179733117421468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,2,power_law_1.2,0.7334527969360352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,1,balanced,0.10478400190671285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,1,balanced,0.10603732864061992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,2,power_law_1.2,1.1522687911987304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,1,balanced,0.10609066486358643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,2,power_law_1.2,2.2690624237060546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,1,balanced,0.10711999734242757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,1,balanced,0.11468799908955891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,1,balanced,0.11577600240707397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,1,balanced,0.11192533373832703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,1,balanced,0.10679999987284343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,1,balanced,0.108815997838974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,1,balanced,0.1111893355846405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,1,balanced,0.10150399804115295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,1,balanced,0.10265599687894185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,1,balanced,0.1072160005569458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,1,balanced,0.10771733522415161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,1,balanced,0.11095999677975972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,1,balanced,0.15256533026695251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,1,balanced,0.14385066429773966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,1,balanced,0.1844373345375061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,1,balanced,0.21394133567810059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,1,balanced,0.28862933317820233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,1,balanced,0.36980799833933514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,1,balanced,0.5297546784083048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,1,balanced,0.6883947054545084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,1,balanced,1.008453369140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,1,balanced,1.3454240163167317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,1,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,1,balanced,0.058602665861447654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,1,balanced,0.07175466914971669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,1,balanced,1.661392052968343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,1,balanced,0.07542933523654938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,1,balanced,0.07733333110809326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,1,balanced,0.07573866844177246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,1,balanced,0.07874133189519246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,1,balanced,0.07702933251857758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,1,balanced,0.046629334489504494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,1,balanced,0.0817493349313736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,1,balanced,0.059157331784566246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,1,balanced,0.07566933333873749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,1,balanced,0.08222933113574982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,1,balanced,2.6489973068237305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,1,balanced,0.0799786647160848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,1,balanced,0.08295466502507527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,1,balanced,0.07514133552710216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,1,balanced,0.07874666651089986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,1,balanced,0.07076266904671986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,1,balanced,0.07650133470694225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,1,balanced,0.08186133205890656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,1,balanced,0.0758186678091685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,1,balanced,0.07128533224264781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,1,balanced,0.08220799763997395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,1,balanced,0.07692266503969829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,1,balanced,0.08566400408744812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,1,balanced,0.0757013310988744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,1,balanced,0.0904960036277771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,1,balanced,0.07601066430409749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,1,balanced,0.11288533608118693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,1,balanced,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,1,balanced,0.11425066987673442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,1,balanced,0.0746666689713796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,1,balanced,0.07531733314196269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,1,balanced,0.1497173309326172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,1,balanced,0.07750399907430013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,1,balanced,0.1757813294728597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,1,balanced,5.25056521097819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,1,balanced,0.07529599964618683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,1,balanced,0.0775786687930425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,1,balanced,0.24078933397928873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,1,balanced,0.11235200365384419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,1,balanced,0.29840532938639325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,1,balanced,0.11154133081436157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,1,balanced,0.13917332887649536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,1,balanced,0.42838935057322186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,1,balanced,0.1671839952468872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,1,balanced,0.21970132986704508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,1,balanced,0.5502453247706095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,1,balanced,0.27179733912150067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,1,balanced,0.019653332730134327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,1,balanced,0.02125866711139679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,1,balanced,0.021727999051411945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,1,balanced,0.38301865259806317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,1,balanced,0.7974666754404703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,1,balanced,0.02272533377011617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,1,balanced,0.024362665911515553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,1,balanced,0.027978666126728058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,1,balanced,0.029882666965325672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,1,balanced,0.49243199825286865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,1,balanced,0.028522667785485584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,1,balanced,0.026154667139053345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,1,balanced,1.0461386839548747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,1,balanced,0.028517333169778187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,1,balanced,0.7054933706919352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,1,balanced,0.030213333666324615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,1,balanced,0.02834133307139079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,1,balanced,0.031109333038330078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,1,balanced,0.032629333436489105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,1,balanced,0.9254453182220459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,1,balanced,1.2946613629659016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,1,balanced,0.04674133161703745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,1,balanced,0.05922133227189382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,1,balanced,0.060975998640060425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,1,balanced,0.048570667703946434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,1,balanced,1.1454986731211345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,1,balanced,0.050586665670077004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,1,balanced,0.063509335120519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,1,balanced,0.07127466797828674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,1,balanced,2.0369973182678223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,1,balanced,0.09738666812578838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,1,balanced,1.798997402191162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,1,balanced,0.11321066816647847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,1,balanced,0.15714133779207864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,1,balanced,0.19829332828521729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,1,balanced,0.284223993619283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,1,balanced,4.027066548665364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,1,balanced,3.5454880396525064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,1,balanced,0.37087468306223553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,1,balanced,0.4551466703414917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,1,balanced,0.7081013520558676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,1,balanced,1.384463946024577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,1,power_law_1.01,0.05630080103874206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,1,power_law_1.01,0.0622655987739563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,1,power_law_1.01,0.07959039807319641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,1,power_law_1.01,0.10930559635162354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,1,power_law_1.01,0.10898560285568237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,1,power_law_1.01,0.11125119924545288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,1,power_law_1.01,0.11489919424057007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,1,power_law_1.01,0.10977280139923096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,1,power_law_1.01,0.11079039573669433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,1,power_law_1.01,0.11030399799346924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,1,power_law_1.01,0.11216000318527222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,1,power_law_1.01,0.11159679889678956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,1,power_law_1.01,0.11088639497756958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,1,power_law_1.01,0.11372159719467163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,1,power_law_1.01,0.1251456022262573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,1,power_law_1.01,0.13162239789962768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,1,power_law_1.01,0.1378432035446167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,1,power_law_1.01,0.16535680294036864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,1,power_law_1.01,0.1804095983505249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,1,power_law_1.01,0.2256767988204956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,1,power_law_1.01,0.2666624069213867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,1,power_law_1.01,0.35797119140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,1,power_law_1.01,0.41982078552246094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,1,power_law_1.01,0.5735040187835694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,1,power_law_1.01,0.7398143768310547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,1,power_law_1.01,1.0602175712585449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,1,power_law_1.01,1.374112033843994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,1,power_law_1.01,1.685958480834961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,1,power_law_1.01,2.665478324890137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,1,power_law_1.01,5.260793685913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,1,power_law_1.01,0.04897280037403107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,1,power_law_1.01,0.05063679814338684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,1,power_law_1.01,0.054604798555374146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,1,power_law_1.01,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,1,power_law_1.01,0.053523200750350955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,1,power_law_1.01,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,1,power_law_1.01,0.07989119887351989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,1,power_law_1.01,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,1,power_law_1.01,0.07816320061683654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,1,power_law_1.01,0.07580800056457519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,1,power_law_1.01,0.07246720194816589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,1,power_law_1.01,0.07621759772300721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,1,power_law_1.01,0.07793279886245727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,1,power_law_1.01,0.07441920042037964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,1,power_law_1.01,0.07519999742507935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,1,power_law_1.01,0.07325440049171447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,1,power_law_1.01,0.07329919934272766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,1,power_law_1.01,0.07508479952812194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,1,power_law_1.01,0.07505279779434204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,1,power_law_1.01,0.07608960270881653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,1,power_law_1.01,0.07589120268821717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,1,power_law_1.01,0.08155519962310791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,1,power_law_1.01,0.08158720135688782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,1,power_law_1.01,0.09111679792404175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,1,power_law_1.01,0.09314560294151306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,1,power_law_1.01,0.10116480588912964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,1,power_law_1.01,0.12070399522781372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,1,power_law_1.01,0.132096004486084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,1,power_law_1.01,0.16517119407653807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,1,power_law_1.01,0.1958400011062622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,1,power_law_1.01,0.2616127967834473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,1,power_law_1.01,0.3168256044387817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,1,power_law_1.01,0.4415103912353516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,1,power_law_1.01,0.5691967964172363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,1,power_law_1.01,0.819052791595459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,1,power_law_1.01,0.018803200125694274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,1,power_law_1.01,1.05862398147583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,1,power_law_1.01,0.019315199553966524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,1,power_law_1.01,1.3133952140808105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,1,power_law_1.01,0.021414400637149812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,1,power_law_1.01,2.0558143615722657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,1,power_law_1.01,0.02131199985742569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,1,power_law_1.01,4.024422454833984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,1,power_law_1.01,0.02149759978055954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,1,power_law_1.01,0.026374399662017822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,1,power_law_1.01,0.026572799682617186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,1,power_law_1.01,0.028377598524093627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,1,power_law_1.01,0.02571519911289215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,1,power_law_1.01,0.02688640058040619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,1,power_law_1.01,0.028172799944877626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,1,power_law_1.01,0.02677760124206543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,1,power_law_1.01,0.029344001412391664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,1,power_law_1.01,0.02942720055580139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,1,power_law_1.01,0.04552960097789764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,1,power_law_1.01,0.05777919888496399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,1,power_law_1.01,0.059910398721694944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,1,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.01,0.04925439953804016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.01,0.06348159909248352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.01,0.07021440267562866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.01,0.0956928014755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.01,0.11216000318527222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.01,0.15587199926376344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.01,0.197107195854187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.01,0.28241920471191406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.01,0.3694591999053955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.01,0.453766393661499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,1,power_law_1.01,0.7071424007415772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,1,power_law_1.01,1.3820416450500488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,1,power_law_1.2,0.05416319966316223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,1,power_law_1.2,0.05783680081367493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,1,power_law_1.2,0.0683135986328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,1,power_law_1.2,0.10997120141983033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,1,power_law_1.2,0.11103999614715576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,1,power_law_1.2,0.11241600513458253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,1,power_law_1.2,0.11011199951171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,1,power_law_1.2,0.1109503984451294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,1,power_law_1.2,0.1096127986907959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,1,power_law_1.2,0.11391999721527099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,1,power_law_1.2,0.11105920076370239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,1,power_law_1.2,0.11427199840545654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,1,power_law_1.2,0.11348479986190796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,1,power_law_1.2,0.11904000043869019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,1,power_law_1.2,0.12766079902648925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,1,power_law_1.2,0.12828799486160278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,1,power_law_1.2,0.1450495958328247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,1,power_law_1.2,0.16240639686584474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,1,power_law_1.2,0.18519680500030516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,1,power_law_1.2,0.22719359397888184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,1,power_law_1.2,0.2533760070800781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,1,power_law_1.2,0.3430527925491333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,1,power_law_1.2,0.4083712100982666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,1,power_law_1.2,0.5617023944854737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,1,power_law_1.2,0.7495808124542236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,1,power_law_1.2,1.0522879600524901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,1,power_law_1.2,1.362112045288086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,1,power_law_1.2,1.6976640701293946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,1,power_law_1.2,2.676326370239258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,1,power_law_1.2,5.272851181030274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,1,power_law_1.2,0.04391680061817169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,1,power_law_1.2,0.04652160108089447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,1,power_law_1.2,0.04970879852771759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,1,power_law_1.2,0.07148799896240235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,1,power_law_1.2,0.07099519968032837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,1,power_law_1.2,0.07319679856300354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,1,power_law_1.2,0.07399680018424988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,1,power_law_1.2,0.0765824019908905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,1,power_law_1.2,0.07846400141716003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,1,power_law_1.2,0.07459840178489685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,1,power_law_1.2,0.07454079985618592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,1,power_law_1.2,0.07861120104789734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,1,power_law_1.2,0.08154879808425904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,1,power_law_1.2,0.08391039967536926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,1,power_law_1.2,0.08858240246772767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,1,power_law_1.2,0.09454079866409301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,1,power_law_1.2,0.10213119983673095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,1,power_law_1.2,0.11884160041809082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,1,power_law_1.2,0.1306048035621643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,1,power_law_1.2,0.16187520027160646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,1,power_law_1.2,0.19319679737091064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,1,power_law_1.2,0.2548288106918335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,1,power_law_1.2,0.31448318958282473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,1,power_law_1.2,0.04563199877738953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,1,power_law_1.2,0.44303359985351565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,1,power_law_1.01,0.07376639842987061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,1,power_law_1.2,0.05242239832878113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,1,power_law_1.2,0.5659584045410156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,1,power_law_1.01,0.07440639734268188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,1,power_law_1.2,0.055731201171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,1,power_law_1.01,0.07685120105743408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,1,power_law_1.2,0.8139904022216797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,1,power_law_1.2,0.07431039810180665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,1,power_law_1.01,0.07883520126342773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,1,power_law_1.2,1.0612544059753417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,1,power_law_1.2,0.07048959732055664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,1,power_law_1.01,0.08423039913177491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,1,power_law_1.2,1.3048447608947753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,1,power_law_1.2,0.0744704008102417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,1,power_law_1.01,0.09207680225372314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,1,power_law_1.2,2.0586048126220704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,1,power_law_1.2,0.07442560195922851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,1,power_law_1.01,0.09262719750404358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,1,power_law_1.2,4.040883255004883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,1,power_law_1.2,0.07484800219535828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,1,power_law_1.01,0.10283520221710205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,1,power_law_1.2,0.07514879703521729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,1,power_law_1.01,0.11662720441818238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,1,power_law_1.2,0.07593600153923034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.01,0.13167359828948974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,1,power_law_1.2,0.0742143988609314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.01,0.15789439678192138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,1,power_law_1.2,0.07885439991950989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.01,0.18318719863891603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,1,power_law_1.2,0.07959039807319641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.01,0.23569281101226808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,1,power_law_1.2,0.08169599771499633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.01,0.2901632070541382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,1,power_law_1.2,0.0948032021522522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.01,0.4032320022583008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,1,power_law_1.2,0.09245439767837524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.01,0.5129600048065186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,1,power_law_1.2,0.09995520114898682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.01,0.7195839881896973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,1,power_law_1.2,0.11261440515518188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.2,0.12785279750823975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.01,0.9332863807678222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.2,0.15432319641113282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.01,1.1500927925109863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.2,0.18495999574661254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,1,power_law_1.01,1.7985599517822266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.2,0.23854079246520996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,1,power_law_1.01,3.5358272552490235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.2,0.2942591905593872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.2,0.4026495933532715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.2,0.5046783924102783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.2,0.7151423931121826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.2,0.9392512321472168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.2,1.1469311714172363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,1,power_law_1.2,1.807596778869629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,1,power_law_1.2,3.5564865112304687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,1,power_law_1.2,0.018822400271892546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,1,power_law_1.2,0.019065600633621217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,1,power_law_1.2,0.021062399446964263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,1,power_law_1.2,0.02128639966249466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,1,power_law_1.2,0.022291199862957002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,1,power_law_1.2,0.026579201221466064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,1,power_law_1.2,0.02571519911289215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,1,power_law_1.2,0.026144000887870788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,1,power_law_1.2,0.024428799748420715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,1,power_law_1.2,0.027321600914001466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,1,power_law_1.2,0.029023998975753786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,1,power_law_1.2,0.02595199942588806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,1,power_law_1.2,0.029849600791931153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,1,power_law_1.2,0.031040000915527343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,1,power_law_1.2,0.04485760033130646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,1,power_law_1.2,0.05862399935722351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,1,power_law_1.2,0.05904639959335327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,1,power_law_1.2,0.047270399332046506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.2,0.049267199635505673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.2,0.06435199975967407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.2,0.070169597864151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.2,0.09536640048027038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.2,0.11375360488891602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.2,0.1557055950164795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.2,0.1993280053138733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.2,0.28475520610809324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.2,0.3696576118469238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.2,0.45236477851867674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,1,power_law_1.2,0.7054272174835206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,1,power_law_1.2,1.3822272300720215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,8,balanced,0.20830933252970377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,8,balanced,0.23690666755040488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,8,balanced,0.23939732710520426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,8,balanced,0.21171200275421143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,8,balanced,0.21377599239349365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,8,balanced,0.22694933414459229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,8,balanced,0.21990933020909628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,8,balanced,0.21497599283854166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,8,balanced,0.20708266894022623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,8,balanced,0.22511466344197592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,8,balanced,0.20908266305923462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,8,balanced,0.2076746622721354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,8,balanced,0.20051199197769165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,8,balanced,0.187226672967275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,8,balanced,0.17990400393803915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,8,balanced,0.17354132731755575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,8,balanced,0.1739413340886434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,8,balanced,0.26239999135335285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,8,balanced,0.1449120044708252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,8,balanced,0.2478613257408142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,8,balanced,0.1507306694984436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,8,balanced,0.3232640027999878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,8,balanced,0.1519200007120768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,8,balanced,0.4276426633199056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,8,balanced,0.15161066253980002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,8,balanced,0.14205867052078247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,8,balanced,0.5776960055033366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,8,balanced,0.14086932937304178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,8,balanced,0.13321066896120706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,8,balanced,0.13206400473912558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,8,balanced,0.7623039881388346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,8,balanced,0.1346506675084432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,8,balanced,0.1304746667544047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,8,balanced,0.1344213287035624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,8,balanced,0.13430933157602945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,8,balanced,0.13095466295878092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,8,balanced,1.0882506370544434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,8,balanced,0.1288159986337026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,8,balanced,0.13367467125256857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,8,balanced,0.12710400422414145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,8,balanced,0.12524267037709555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,8,balanced,1.4364107449849446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,8,balanced,0.11922666430473328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,8,balanced,0.1316373348236084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,8,balanced,0.1211946705977122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,8,balanced,0.1209440032641093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,8,balanced,0.12483200430870056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,8,balanced,0.11945600310961406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,8,balanced,0.12363200386365254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,8,balanced,2.108677387237549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,8,balanced,0.12553067008654276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,8,balanced,0.11813867092132568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,8,balanced,0.1160533328851064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,8,balanced,0.12288533647855122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,8,balanced,0.11807466546694438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,8,balanced,0.1735466718673706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,8,balanced,0.114847997824351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,8,balanced,2.7882614135742188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,8,balanced,0.16622933745384216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,8,balanced,0.1207360029220581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,8,balanced,0.2188106576601664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,8,balanced,0.10341866811116536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,8,balanced,0.1055466632048289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,8,balanced,0.2784159978230794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,8,balanced,0.09671466549237569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,8,balanced,3.4306933085123696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,8,balanced,0.37165331840515137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,8,balanced,0.1618826687335968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,8,balanced,0.13011733690897623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,8,balanced,0.4785706599553426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,8,balanced,0.1639359990755717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,8,balanced,0.19750932852427164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,8,balanced,0.6668213208516439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,8,balanced,0.26654932896296185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,8,balanced,5.398746490478516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,8,balanced,0.33562131722768146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,8,balanced,0.8668533166249593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,8,balanced,0.46005332469940186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,8,balanced,1.2395306428273518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,8,balanced,0.5916906595230103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,8,balanced,1.627343972524007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,8,balanced,0.8459786574045817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,8,balanced,10.321050643920898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,8,balanced,1.1046453317006428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,8,balanced,1.9642186164855957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,8,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,8,balanced,0.04889066517353058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,8,balanced,1.3603092829386394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,8,balanced,0.04715733230113983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,8,balanced,0.04859733581542969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,8,balanced,0.04896533489227295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,8,balanced,0.05504000186920166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,8,balanced,0.08964266379674275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,8,balanced,2.929258664449056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,8,balanced,0.08982400099436443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,8,balanced,0.09645332892735799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,8,balanced,0.08956799904505412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,8,balanced,2.1156586011250815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,8,balanced,0.0777706652879715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,8,balanced,0.13800533612569174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,8,balanced,0.1306826670964559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,8,balanced,0.10167466600735982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,8,balanced,0.1449066698551178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,8,balanced,0.124208003282547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,8,balanced,0.15930666526158652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,8,balanced,0.209007998307546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,8,balanced,0.267578661441803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,8,balanced,0.3755840063095093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,8,balanced,0.49406933784484863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,8,balanced,4.142271995544434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,8,balanced,0.7064266999562582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,8,balanced,5.460645039876302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,8,balanced,0.9319679737091064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,8,balanced,1.3697919845581055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,8,balanced,1.8042346636454265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,8,balanced,2.676762580871582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,8,balanced,3.5501279830932617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,8,balanced,4.422368049621582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,8,balanced,7.047909418741862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,8,power_law_1.01,0.23538560867309571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,8,power_law_1.01,0.23557760715484619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,8,balanced,14.036677042643229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,8,power_law_1.01,0.23411200046539307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,8,power_law_1.01,0.2340928077697754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,8,power_law_1.01,0.2224191904067993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,8,power_law_1.01,0.21478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,8,power_law_1.01,0.2100224018096924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,8,power_law_1.01,0.19619840383529663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,8,power_law_1.01,0.19329919815063476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,8,power_law_1.01,0.17761919498443604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,8,power_law_1.01,0.17227519750595094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,8,power_law_1.01,0.18837120532989501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,8,power_law_1.01,0.2268224000930786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,8,power_law_1.01,0.24330239295959472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,8,power_law_1.01,0.3316927909851074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,8,power_law_1.01,0.3447423934936523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,8,power_law_1.01,0.41907200813293455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,8,power_law_1.01,0.5361152172088623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,8,power_law_1.01,0.689734411239624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,8,power_law_1.01,0.9895872116088867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,8,power_law_1.01,1.3191743850708009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,8,power_law_1.01,2.0493696212768553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,8,power_law_1.01,2.3281728744506838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,8,power_law_1.01,3.55230712890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,8,power_law_1.01,4.790534210205078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,8,power_law_1.01,7.576947021484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,8,power_law_1.01,9.439148712158204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,8,power_law_1.01,0.12777600288391114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,8,power_law_1.01,0.14807679653167724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,8,power_law_1.01,0.1201856017112732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,8,power_law_1.01,0.15004160404205322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,8,power_law_1.01,12.13544921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,8,power_law_1.01,0.1318336009979248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,8,power_law_1.01,0.1373568058013916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,8,power_law_1.01,0.12176640033721924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,8,power_law_1.01,0.13949439525604249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,8,power_law_1.01,17.107270812988283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,8,power_law_1.01,0.12279679775238037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,8,power_law_1.01,0.14279680252075194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,8,power_law_1.01,0.11999360322952271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,8,power_law_1.01,0.13408639430999755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,8,power_law_1.01,0.11445759534835816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,8,power_law_1.01,0.13121919631958007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,8,power_law_1.01,32.65138549804688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,8,power_law_1.01,0.12364799976348877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,8,power_law_1.01,0.11601920127868652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,8,power_law_1.01,0.12035839557647705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,8,power_law_1.01,0.11022080183029175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,8,power_law_1.01,0.11326080560684204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,8,power_law_1.01,0.09822720289230347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,8,power_law_1.01,0.1102336049079895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,8,power_law_1.01,0.0937279999256134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,8,power_law_1.01,0.14805760383605956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,8,power_law_1.01,0.13259520530700683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,8,power_law_1.01,0.14740480184555055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,8,power_law_1.01,0.1347584009170532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,8,power_law_1.01,0.15354880094528198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,8,power_law_1.01,0.1381183981895447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,8,power_law_1.01,0.18982399702072145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,8,power_law_1.01,0.17336959838867189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,8,power_law_1.01,0.19339519739151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,8,power_law_1.01,0.15003520250320435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,8,power_law_1.01,0.2511104106903076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,8,power_law_1.01,0.19644800424575806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,8,power_law_1.01,0.33935999870300293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,8,power_law_1.01,0.25178239345550535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,8,power_law_1.01,0.382751989364624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.01,0.30208001136779783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.01,0.3992000102996826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,8,power_law_1.01,0.5528255939483643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.01,0.5093376159667968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,8,power_law_1.01,0.6332032203674316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.01,0.7005184173583985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,8,power_law_1.01,0.9466048240661621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.01,0.9791040420532227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,8,power_law_1.01,1.3265088081359864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.01,1.438316822052002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,8,power_law_1.01,1.9458879470825194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.01,1.8069375991821288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,8,power_law_1.01,2.282624053955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.01,2.6639360427856444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,8,power_law_1.01,3.6029312133789064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.01,3.4033279418945312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,8,power_law_1.01,4.615673446655274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,8,power_law_1.01,4.605843353271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,8,power_law_1.01,5.495923233032227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,8,power_law_1.01,6.811795043945312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,8,power_law_1.01,7.691558074951172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,8,power_law_1.01,12.93489990234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,8,power_law_1.01,16.157408142089842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,8,power_law_1.01,0.045407998561859134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,8,power_law_1.01,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,8,power_law_1.01,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,8,power_law_1.01,0.047884801030159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,8,power_law_1.01,0.04819200038909912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,8,power_law_1.01,0.05377920269966126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,8,power_law_1.01,0.0856383979320526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,8,power_law_1.01,0.08889600038528442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,8,power_law_1.01,0.09167360067367554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,8,power_law_1.01,0.08572160005569458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,8,power_law_1.01,0.07538560032844543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,8,power_law_1.01,0.13551360368728638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,8,power_law_1.01,0.12823679447174072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,8,power_law_1.01,0.09961599707603455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,8,power_law_1.01,0.14298880100250244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,8,power_law_1.01,0.12257280349731445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,8,power_law_1.01,0.15738879442214965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,8,power_law_1.01,0.20526719093322754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.01,0.2651648044586182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.01,0.37433600425720215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.01,0.491871976852417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.01,0.7062464237213135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.01,0.9283007621765137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.01,1.366316795349121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.01,1.8060287475585937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.01,2.6854143142700195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.01,3.5490367889404295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,8,power_law_1.01,4.428518295288086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,8,power_law_1.01,7.045894622802734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,8,power_law_1.01,14.054701232910157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,8,power_law_1.2,0.22904319763183595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,8,power_law_1.2,0.23680639266967773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,8,power_law_1.2,0.2352895975112915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,8,power_law_1.2,0.2359999895095825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,8,power_law_1.2,0.228985595703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,8,power_law_1.2,0.2090303897857666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,8,power_law_1.2,0.20290560722351075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,8,power_law_1.2,0.18246400356292725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,8,power_law_1.2,0.18215680122375488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,8,power_law_1.2,0.1764032006263733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,8,power_law_1.2,0.17146240472793578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,8,power_law_1.2,0.23159680366516114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,8,power_law_1.2,0.21006081104278565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,8,power_law_1.2,0.23827199935913085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,8,power_law_1.2,0.30638720989227297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,8,power_law_1.2,0.31755518913269043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,8,power_law_1.2,0.3665855884552002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,8,power_law_1.2,0.5627200126647949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,8,power_law_1.2,0.7293119907379151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,8,power_law_1.2,1.0619071960449218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,8,power_law_1.2,1.2725055694580079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,8,power_law_1.2,0.15073920488357545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,8,power_law_1.2,0.12602880001068115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,8,power_law_1.2,0.13201919794082642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,8,power_law_1.2,1.9495424270629882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,8,power_law_1.2,0.1272063970565796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,8,power_law_1.2,0.13216639757156373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,8,power_law_1.2,2.4946367263793947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,8,power_law_1.2,0.11953279972076417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,8,power_law_1.2,0.13937920331954956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,8,power_law_1.2,3.922515106201172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,8,power_law_1.2,0.13623679876327516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,8,power_law_1.2,0.12694400548934937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,8,power_law_1.2,5.177824020385742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,8,power_law_1.2,0.1348031997680664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,8,power_law_1.2,0.1261888027191162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,8,power_law_1.2,0.1264896035194397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,8,power_law_1.2,0.122489595413208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,8,power_law_1.2,7.8059837341308596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,8,power_law_1.2,0.1297600030899048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,8,power_law_1.2,0.11692160367965698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,8,power_law_1.2,10.014157104492188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,8,power_law_1.2,0.11706240177154541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,8,power_law_1.2,0.11575679779052735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,8,power_law_1.2,0.12158080339431762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,8,power_law_1.2,0.11292799711227416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,8,power_law_1.2,11.671616363525391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,8,power_law_1.2,0.11048959493637085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,8,power_law_1.2,0.10654079914093018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,8,power_law_1.2,0.14568959474563598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,8,power_law_1.2,0.0930624008178711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,8,power_law_1.2,0.15699199438095093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,8,power_law_1.2,0.11934080123901367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,8,power_law_1.2,17.740864562988282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,8,power_law_1.2,0.13377920389175416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,8,power_law_1.2,0.15160319805145264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,8,power_law_1.2,0.18272639513015748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,8,power_law_1.2,0.12917759418487548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,8,power_law_1.2,0.19522559642791748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,8,power_law_1.2,0.15750399827957154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,8,power_law_1.2,32.365133666992186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,8,power_law_1.2,0.23196160793304443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,8,power_law_1.2,0.15589120388031005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,8,power_law_1.2,0.31182079315185546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,8,power_law_1.2,0.17793920040130615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,8,power_law_1.2,0.4219327926635742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,8,power_law_1.2,0.24885120391845703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,8,power_law_1.2,0.5853504180908203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.2,0.2928447961807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,8,power_law_1.2,0.7321599960327149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.2,0.4066688060760498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,8,power_law_1.2,1.0330559730529785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.2,0.5045631885528564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.2,0.7248256206512451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,8,power_law_1.2,1.3950016021728515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.2,0.9291199684143067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,8,power_law_1.2,1.8846656799316406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.2,1.43057279586792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,8,power_law_1.2,2.4223487854003904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.2,1.7942207336425782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,8,power_law_1.2,3.4432960510253907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.2,2.672524833679199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,8,power_law_1.2,4.345862579345703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.2,3.7034751892089846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,8,power_law_1.2,5.641004943847657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,8,power_law_1.2,4.307833480834961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,8,power_law_1.2,8.517823791503906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,8,power_law_1.2,7.36087646484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,8,power_law_1.2,16.502951049804686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,8,power_law_1.2,14.583219909667969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,8,power_law_1.2,0.04726400077342987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,8,power_law_1.2,0.046937599778175354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,8,power_law_1.2,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,8,power_law_1.2,0.04761599898338318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,8,power_law_1.2,0.048640000820159915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,8,power_law_1.2,0.053324800729751584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,8,power_law_1.2,0.08780800104141236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,8,power_law_1.2,0.08754559755325317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,8,power_law_1.2,0.09575679898262024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,8,power_law_1.2,0.08760960102081299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,8,power_law_1.2,0.0763584017753601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,8,power_law_1.2,0.1401792049407959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,8,power_law_1.2,0.12828799486160278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,8,power_law_1.2,0.09960319995880126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,8,power_law_1.2,0.14421759843826293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,8,power_law_1.2,0.12136319875717164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,8,power_law_1.2,0.15854079723358155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,8,power_law_1.2,0.2055488109588623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.2,0.2663935899734497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.2,0.37526400089263917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.2,0.49233279228210447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.2,0.7059135913848877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.2,0.932198429107666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.2,1.365516757965088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.2,1.801100730895996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.2,2.67574405670166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.2,3.5510913848876955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,8,power_law_1.2,4.420761489868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,8,power_law_1.2,7.043218994140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,8,power_law_1.2,14.051834106445312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,4,balanced,0.12496532996495564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,4,balanced,0.14296000202496847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,4,balanced,0.176362673441569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,4,balanced,0.19153600931167603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,4,balanced,0.18850666284561157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,4,balanced,0.18563199043273926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,4,balanced,0.1843573252360026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,4,balanced,0.1841599941253662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,4,balanced,0.19203199942906699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,4,balanced,0.1827039917310079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,4,balanced,0.17882132530212402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,4,balanced,0.16819200913111368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,4,balanced,0.16219733158747354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,4,balanced,0.1601599951585134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,4,balanced,0.15678399801254272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,4,balanced,0.0835093359152476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,4,balanced,0.15028799573580423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,4,balanced,0.08715200424194336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,4,balanced,0.08457066615422566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,4,balanced,0.15178133050600687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,4,balanced,0.1306773324807485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,4,balanced,0.07585066556930542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,4,balanced,0.2600160042444865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,4,balanced,0.1167093316713969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,4,balanced,0.12735467155774435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,4,balanced,0.12058132886886597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,4,balanced,0.2425653338432312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,4,balanced,0.12433066964149475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,4,balanced,0.11557333668073018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,4,balanced,0.30300267537434894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,4,balanced,0.11192533373832703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,4,balanced,0.11814399560292561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,4,balanced,0.11486933628718059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,4,balanced,0.12044800321261089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,4,balanced,0.38858668009440106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,4,balanced,0.12197333574295044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,4,balanced,0.12745066483815512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,4,balanced,0.12172800302505493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,4,balanced,0.5391519864400228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,4,balanced,0.11449600259462993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,4,balanced,0.12063999970753987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,4,balanced,0.11814399560292561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,4,balanced,0.706117312113444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,4,balanced,0.11924266815185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,4,balanced,0.1111306647459666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,4,balanced,0.10748799641927083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,4,balanced,0.10716799894968669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,4,balanced,0.11648533741633098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,4,balanced,0.11152533690134685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,4,balanced,1.0063786506652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,4,balanced,0.11500266194343567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,4,balanced,0.1092800001303355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,4,balanced,0.10430399576822917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,4,balanced,0.10582933823267619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,4,balanced,0.10195199648539226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,4,balanced,1.2949972947438557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,4,balanced,0.10846933722496033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,4,balanced,0.09823466340700786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,4,balanced,0.1663040022055308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,4,balanced,0.0950879951318105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,4,balanced,0.1628213326136271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,4,balanced,1.878752072652181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,4,balanced,0.14522666732470194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,4,balanced,0.20198400815327963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,4,balanced,0.12086932857831319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,4,balanced,0.2444960077603658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,4,balanced,0.15446399648984274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,4,balanced,2.3763839403788247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,4,balanced,0.32608532905578613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,4,balanced,0.18889067570368448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,4,balanced,0.4179519812266032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,4,balanced,0.24800533056259155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,4,balanced,2.898144086201986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,4,balanced,0.5704426765441895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,4,balanced,0.3117973407109578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,4,balanced,0.7356533209482828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,4,balanced,0.4378559986750285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,4,balanced,0.5613280137379965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,4,balanced,1.051514705022176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,4,balanced,4.312896092732747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,4,balanced,0.8142666816711426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,4,balanced,1.3656907081604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,4,balanced,1.0689226786295574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,4,balanced,1.67523193359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,4,balanced,8.12931760152181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,4,balanced,1.3215146859486897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,4,balanced,2.6185760498046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,4,balanced,2.0880373318990073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,4,balanced,0.02924266705910365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,4,balanced,0.028778667251269024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,4,balanced,5.099664052327474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,4,balanced,0.028192001084486645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,4,balanced,0.033039999504884086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,4,balanced,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,4,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,4,balanced,0.04855999847253164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,4,balanced,0.04104000081618627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,4,balanced,0.08194666604201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,4,balanced,0.07995733122030894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,4,balanced,0.08036266764005025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,4,balanced,0.0867146650950114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,4,balanced,4.1383358637491865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,4,balanced,0.08210666477680206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,4,balanced,0.06332799792289734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,4,balanced,0.09434666236241658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,4,balanced,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,4,balanced,0.09524266918500264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,4,balanced,0.11699199676513672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,4,balanced,0.1421066621939341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,4,balanced,0.19471466541290283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,4,balanced,0.24714666604995728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,4,balanced,0.3464053471883138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,4,balanced,0.4461173216501872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,4,balanced,0.6434719959894816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,4,balanced,0.840949296951294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,4,balanced,1.2332159678141277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,4,balanced,1.626245339711507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,4,balanced,2.0226613680521646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,4,power_law_1.01,0.16427520513534546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,4,balanced,3.1998027165730796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,4,power_law_1.01,0.16012799739837646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,4,power_law_1.01,0.16647039651870726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,4,power_law_1.01,0.18703999519348144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,4,power_law_1.01,0.1938752055168152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,4,power_law_1.01,0.1870784044265747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,4,power_law_1.01,0.18536959886550902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,4,power_law_1.01,0.1864575982093811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,4,power_law_1.01,0.17734400033950806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,4,power_law_1.01,0.17566720247268677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,4,power_law_1.01,0.16641279458999633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,4,power_law_1.01,0.18873599767684937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,4,power_law_1.01,0.18384640216827391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,4,power_law_1.01,0.1879487991333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,4,balanced,6.344730377197266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,4,power_law_1.01,0.24971520900726318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,4,power_law_1.01,0.2697664022445679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,4,power_law_1.01,0.0996288001537323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,4,power_law_1.01,0.28852479457855223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,4,power_law_1.01,0.11466879844665527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,4,power_law_1.01,0.3674304008483887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,4,power_law_1.01,0.49680638313293457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,4,power_law_1.01,0.11683199405670167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,4,power_law_1.01,0.11917439699172974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,4,power_law_1.01,0.6776959896087646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,4,power_law_1.01,0.11422079801559448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,4,power_law_1.01,0.8610943794250489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,4,power_law_1.01,0.1170240044593811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,4,power_law_1.01,1.2674688339233398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,4,power_law_1.01,0.1171839952468872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,4,power_law_1.01,1.540403175354004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,4,power_law_1.01,0.12014080286026001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,4,power_law_1.01,2.2696895599365234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,4,power_law_1.01,0.11802879571914673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,4,power_law_1.01,3.0939968109130858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,4,power_law_1.01,0.113811194896698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,4,power_law_1.01,4.133574295043945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,4,power_law_1.01,0.11047040224075318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,4,power_law_1.01,5.7252544403076175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,4,power_law_1.01,0.12480640411376953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,4,power_law_1.01,0.12431999444961547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,4,power_law_1.01,6.022905731201172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,4,power_law_1.01,0.131660795211792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,4,power_law_1.01,9.77462387084961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,4,power_law_1.01,0.1540544033050537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,4,power_law_1.01,0.15349119901657104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,4,power_law_1.01,0.09822720289230347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,4,power_law_1.01,0.1752832055091858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,4,power_law_1.01,22.514738464355467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,4,power_law_1.01,0.10559359788894654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,4,power_law_1.01,0.24896640777587892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,4,power_law_1.01,0.10513919591903687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,4,power_law_1.01,0.27127039432525635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,4,power_law_1.01,0.11272959709167481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,4,power_law_1.01,0.35031681060791015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,4,power_law_1.01,0.11311359405517578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,4,power_law_1.01,0.48931198120117186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,4,power_law_1.01,0.11545599699020385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,4,power_law_1.01,0.5868671894073486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,4,power_law_1.01,0.11598080396652222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,4,power_law_1.01,0.7925568103790284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,4,power_law_1.01,0.11303679943084717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,4,power_law_1.01,1.0224960327148438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,4,power_law_1.01,0.10967040061950684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,4,power_law_1.01,1.643142318725586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,4,power_law_1.01,0.10865919589996338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,4,power_law_1.01,2.5392383575439452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,4,power_law_1.01,0.10343040227890014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,4,power_law_1.01,2.966975975036621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,4,power_law_1.01,0.10885119438171387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,4,power_law_1.01,0.12524800300598143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,4,power_law_1.01,3.4066497802734377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,4,power_law_1.01,0.12469760179519654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,4,power_law_1.01,4.620377731323242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,4,power_law_1.01,0.13472000360488892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,4,power_law_1.01,0.14123519659042358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,4,power_law_1.01,12.620448303222656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,4,power_law_1.01,0.150873601436615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,4,power_law_1.01,0.1931071996688843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.01,0.2149183988571167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.01,0.2861119985580444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.01,0.33678719997406004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.01,0.46541438102722166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.01,0.6372159957885742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.01,0.8182016372680664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.01,1.2374272346496582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.01,2.1642688751220702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.01,2.657241630554199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,4,power_law_1.01,2.92674560546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,4,power_law_1.01,5.205446243286133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,4,power_law_1.01,11.604486083984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,4,power_law_1.01,0.02743679881095886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,4,power_law_1.01,0.026086398959159852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,4,power_law_1.01,0.026393601298332216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,4,power_law_1.01,0.03108479976654053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,4,power_law_1.01,0.0322816014289856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,4,power_law_1.01,0.03238399922847748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,4,power_law_1.01,0.04774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,4,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,4,power_law_1.01,0.07948799729347229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,4,power_law_1.01,0.07760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,4,power_law_1.01,0.07894399762153625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,4,power_law_1.01,0.0865343987941742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,4,power_law_1.01,0.0815168023109436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,4,power_law_1.2,0.16301440000534057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,4,power_law_1.01,0.05920640230178833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,4,power_law_1.2,0.18384000062942504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,4,power_law_1.01,0.09315840005874634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,4,power_law_1.2,0.14677120447158815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,4,power_law_1.01,0.07193599939346314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,4,power_law_1.2,0.19994239807128905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,4,power_law_1.01,0.09226880073547364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,4,power_law_1.2,0.1931007981300354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,4,power_law_1.01,0.11433600187301636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,4,power_law_1.2,0.19215999841690062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.01,0.13833600282669067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,4,power_law_1.2,0.18332159519195557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.01,0.19445120096206664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,4,power_law_1.2,0.1765120029449463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.01,0.24498560428619384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,4,power_law_1.2,0.1773695945739746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.01,0.34622719287872317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,4,power_law_1.2,0.16785279512405396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.01,0.44460158348083495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,4,power_law_1.2,0.16579840183258057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.01,0.6403584003448486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,4,power_law_1.2,0.19367680549621583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.01,0.838163185119629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,4,power_law_1.2,0.2014591932296753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.01,1.238368034362793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,4,power_law_1.2,0.1982591986656189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.01,1.6316287994384766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,4,power_law_1.2,0.1032639980316162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,4,power_law_1.2,0.2773888111114502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,4,power_law_1.01,2.0173952102661135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,4,power_law_1.2,0.1055616021156311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,4,power_law_1.2,0.3010175943374634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,4,power_law_1.01,3.198099136352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,4,power_law_1.2,0.09596160054206848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,4,power_law_1.2,0.3122495889663696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,4,power_law_1.2,0.11750400066375732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,4,power_law_1.2,0.3989696025848389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,4,power_law_1.01,6.338431930541992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,4,power_law_1.2,0.12023680210113526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,4,power_law_1.2,0.48018560409545896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,4,power_law_1.2,0.11875840425491332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,4,power_law_1.2,0.699948787689209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,4,power_law_1.2,0.11482239961624145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,4,power_law_1.2,0.7436800003051758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,4,power_law_1.2,0.11592960357666016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,4,power_law_1.2,1.4010432243347168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,4,power_law_1.2,0.11618560552597046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,4,power_law_1.2,1.7798656463623046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,4,power_law_1.2,0.1118783950805664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,4,power_law_1.2,2.8997823715209963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,4,power_law_1.2,0.10600320100784302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,4,power_law_1.2,3.2257728576660156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,4,power_law_1.2,0.13466880321502686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,4,power_law_1.2,4.637158584594727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,4,power_law_1.2,0.14444799423217775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,4,power_law_1.2,0.12386560440063477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,4,power_law_1.2,5.395487976074219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,4,power_law_1.2,0.1683519959449768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,4,power_law_1.2,7.656179046630859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,4,power_law_1.2,0.1601472020149231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,4,power_law_1.2,0.18348159790039062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,4,power_law_1.2,13.500607299804688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,4,power_law_1.2,0.23864960670471191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,4,power_law_1.2,0.2985663890838623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,4,power_law_1.2,0.4077439785003662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,4,power_law_1.2,20.970681762695314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,4,power_law_1.2,0.4103360176086426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,4,power_law_1.2,0.6052544116973877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,4,power_law_1.2,0.8942399978637695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,4,power_law_1.2,1.3882880210876465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,4,power_law_1.2,0.09825279712677001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,4,power_law_1.2,1.814214324951172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,4,power_law_1.2,0.10344959497451782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,4,power_law_1.2,0.11745280027389526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,4,power_law_1.2,2.2043264389038084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,4,power_law_1.2,0.1152959942817688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,4,power_law_1.2,3.8205055236816405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,4,power_law_1.2,0.11366399526596069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,4,power_law_1.2,3.507980728149414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,4,power_law_1.2,0.11274880170822144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,4,power_law_1.2,0.11253759860992432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,4,power_law_1.2,6.873260498046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,4,power_law_1.2,0.11338239908218384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,4,power_law_1.2,14.721913146972657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,4,power_law_1.2,0.10871039628982544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,4,power_law_1.2,0.11208319664001465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,4,power_law_1.2,0.10490880012512208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,4,power_law_1.2,0.11400959491729737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,4,power_law_1.2,0.12060799598693847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,4,power_law_1.2,0.1208448052406311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,4,power_law_1.2,0.1463487982749939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,4,power_law_1.2,0.15057920217514037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,4,power_law_1.2,0.1586176037788391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,4,power_law_1.2,0.18622080087661744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.2,0.22435200214385986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.2,0.31784958839416505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.2,0.3360127925872803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.2,0.5658751964569092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.2,0.8259008407592774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.2,1.1719615936279297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.2,1.5715200424194335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.2,2.2257663726806642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.2,2.8576000213623045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,4,power_law_1.2,3.66890869140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,4,power_law_1.2,5.822406387329101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,4,power_law_1.2,12.559449768066406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,4,power_law_1.2,0.026745599508285523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,4,power_law_1.2,0.025990399718284606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,4,power_law_1.2,0.026111999154090883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,4,power_law_1.2,0.030348798632621764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,4,power_law_1.2,0.0310591995716095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,4,power_law_1.2,0.03278720080852508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,4,power_law_1.2,0.046163201332092285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,4,power_law_1.2,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,4,power_law_1.2,0.07979519963264466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,4,power_law_1.2,0.07820799946784973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,4,power_law_1.2,0.07837439775466919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,4,power_law_1.2,0.08375679850578308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,4,power_law_1.2,0.08020480275154114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,4,power_law_1.2,0.061273598670959474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,4,power_law_1.2,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,4,power_law_1.2,0.07784960269927979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,4,power_law_1.2,0.0946175992488861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,4,power_law_1.2,0.11426559686660767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.2,0.14017280340194702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.2,0.19496959447860718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,2,balanced,0.08943466345469157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.2,0.24746880531311036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,2,balanced,0.11186133821805318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.2,0.34611198902130125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,2,balanced,0.15923733512560526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.2,0.4443520069122314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,2,balanced,0.18200532595316568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.2,0.6416831970214844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,2,balanced,0.17935466766357422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.2,0.8415552139282226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,2,balanced,0.1663040022055308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,2,balanced,0.17331733306248984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.2,1.2322175979614258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,2,balanced,0.1678719917933146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.2,1.6289728164672852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,2,balanced,0.1939199964205424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,4,power_law_1.2,2.0232383728027346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,2,balanced,0.1725013256072998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,2,balanced,0.06889600058396657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,4,power_law_1.2,3.1988031387329103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,2,balanced,0.17246399323145548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,2,balanced,0.07998399933179219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,2,balanced,0.16405333081881204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,4,power_law_1.2,6.3464000701904295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,2,balanced,0.09938133756319682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,2,balanced,0.17358932892481485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,2,balanced,0.1097813347975413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,2,balanced,0.14917866388956705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,2,balanced,0.11618133385976155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,2,balanced,0.15024532874425253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,2,balanced,0.10770133137702942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,2,balanced,0.15212266643842062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,2,balanced,0.11135466893513997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,2,balanced,0.155541330575943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,2,balanced,0.10798933108647664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,2,balanced,0.2318026622136434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,2,balanced,0.12195199728012085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,2,balanced,0.221178670724233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,2,balanced,0.10677867134412129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,2,balanced,0.1102186640103658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,2,balanced,0.2789066632588704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,2,balanced,0.10774399836858113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,2,balanced,0.341541330019633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,2,balanced,0.10599467158317566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,2,balanced,0.10434666275978088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,2,balanced,0.46851201852162677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,2,balanced,0.10801600416501363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,2,balanced,0.10659199953079224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,2,balanced,0.5921066602071127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,2,balanced,0.11123200257619222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,2,balanced,0.15475733081499735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,2,balanced,0.8265813191731771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,2,balanced,0.15005866686503092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,2,balanced,0.18878400325775146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,2,balanced,1.0681119759877522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,2,balanced,0.22196267048517862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,2,balanced,0.05553600192070007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,2,balanced,0.30451732873916626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,2,balanced,0.0757173349459966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,2,balanced,1.558538595835368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,2,balanced,0.11869866649309795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,2,balanced,0.10384533802668254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,2,balanced,0.38556798299153644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,2,balanced,0.10404800375302632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,2,balanced,0.10528000195821126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,2,balanced,2.025514602661133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,2,balanced,0.5515360037485758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,2,balanced,0.11471466223398845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,2,balanced,0.1223360002040863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,2,balanced,0.11268799503644307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,2,balanced,0.7112320264180502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,2,balanced,0.11272000273068745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,2,balanced,0.10597333312034607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,2,balanced,2.498703956604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,2,balanced,0.10967466235160828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,2,balanced,0.10010133186976115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,2,balanced,1.0428799788157146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,2,balanced,0.09980266292889912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,2,balanced,0.09814400474230449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,2,balanced,0.09610666831334432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,2,balanced,0.09241599837938945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,2,balanced,3.9663572311401367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,2,balanced,0.14761066436767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,2,balanced,1.3759840329488118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,2,balanced,0.12904000282287598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,2,balanced,0.15738133589426676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,2,balanced,0.1941439906756083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,2,balanced,0.258517324924469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,2,balanced,1.691391944885254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,2,balanced,0.324069341023763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,2,balanced,0.45763734976450604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,2,balanced,7.874538421630859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,2,balanced,0.5862559874852499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,2,balanced,2.645770708719889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,2,balanced,0.8500693639119467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,2,balanced,1.1236000061035156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,2,balanced,1.382373332977295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,2,balanced,5.25271479288737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,2,balanced,2.178565343221029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,2,balanced,0.0234400009115537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,2,balanced,0.0322080006202062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,2,balanced,0.02492800106604894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,2,balanced,0.02610666553179423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,2,balanced,0.0322026660044988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,2,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,2,balanced,0.030991998811562855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,2,balanced,0.03310399999221166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,2,balanced,0.035775999228159584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,2,balanced,4.332069396972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,2,balanced,0.03472533325354258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,2,balanced,0.03724266588687897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,2,balanced,0.07387733459472656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,2,balanced,0.07566933333873749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,2,balanced,0.07977066437403361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,2,balanced,0.06764266888300578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,2,balanced,0.05508799850940704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,2,balanced,0.06046933432420095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,2,balanced,0.075914666056633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,2,balanced,0.09225066502888997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,2,balanced,0.12293866276741028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,2,balanced,0.14613333344459534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,2,balanced,0.20453866322835287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,2,balanced,0.25835732618967694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,2,balanced,0.3678026596705119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,2,balanced,0.4780426820119222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,2,balanced,0.6948800086975098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,2,balanced,0.910213311513265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,2,power_law_1.01,0.10044159889221191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,2,balanced,1.1285759607950847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,2,power_law_1.01,0.12548480033874512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,2,power_law_1.01,0.13825279474258423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,2,power_law_1.01,0.1821951985359192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,2,power_law_1.01,0.17634559869766236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,2,power_law_1.01,0.1778048038482666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,2,power_law_1.01,0.18615039587020873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,2,power_law_1.01,0.16856319904327394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,2,balanced,1.7771520614624023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,2,power_law_1.01,0.17019519805908204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,2,power_law_1.01,0.17076480388641357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,2,power_law_1.01,0.16733440160751342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,2,power_law_1.01,0.1727231979370117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,2,power_law_1.01,0.18488320112228393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,2,power_law_1.01,0.18028160333633422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,2,power_law_1.01,0.1932031989097595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,2,power_law_1.01,0.22748160362243652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,2,power_law_1.01,0.2528575897216797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,2,power_law_1.01,0.27733759880065917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,2,power_law_1.01,0.3278271913528442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,2,power_law_1.01,0.40628480911254883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,2,power_law_1.01,0.5206208229064941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,2,balanced,3.5086612701416016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,2,power_law_1.01,0.6875199794769287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,2,power_law_1.01,0.8590399742126464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,2,power_law_1.01,0.0723583996295929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,2,power_law_1.01,1.27258243560791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,2,power_law_1.01,0.07377920150756836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,2,power_law_1.01,0.08725119829177856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,2,power_law_1.01,1.8176319122314453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,2,power_law_1.01,0.10865919589996338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,2,power_law_1.01,2.4622528076171877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,2,power_law_1.01,0.11300480365753174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,2,power_law_1.01,3.1125823974609377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,2,power_law_1.01,0.11459840536117553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,2,power_law_1.01,0.07148799896240235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,2,power_law_1.01,4.046694564819336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,2,power_law_1.01,0.11221760511398315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,2,power_law_1.01,0.10099200010299683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,2,power_law_1.01,0.11072640419006348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,2,power_law_1.01,5.846035385131836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,2,power_law_1.01,0.08962560296058655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,2,power_law_1.01,0.1091264009475708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,2,power_law_1.01,0.1023743987083435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,2,power_law_1.01,14.134617614746094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,2,power_law_1.01,0.11138559579849243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,2,power_law_1.01,0.11117440462112427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,2,power_law_1.01,0.1080064058303833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,2,power_law_1.01,0.11437439918518066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,2,power_law_1.01,0.11950080394744873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,2,power_law_1.01,0.10485759973526002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,2,power_law_1.01,0.11978880167007447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,2,power_law_1.01,0.11167999505996704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,2,power_law_1.01,0.12426879405975341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,2,power_law_1.01,0.10670080184936523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,2,power_law_1.01,0.14034559726715087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,2,power_law_1.01,0.10483200550079345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,2,power_law_1.01,0.13365119695663452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,2,power_law_1.01,0.1056704044342041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,2,power_law_1.01,0.14495999813079835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,2,power_law_1.01,0.11021440029144287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,2,power_law_1.01,0.1871999979019165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,2,power_law_1.01,0.11041280031204223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,2,power_law_1.01,0.22438399791717528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,2,power_law_1.01,0.11079679727554322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,2,power_law_1.01,0.27254400253295896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,2,power_law_1.01,0.1316799998283386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,2,power_law_1.01,0.3166591882705688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,2,power_law_1.01,0.13259520530700683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,2,power_law_1.01,0.5135424137115479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,2,power_law_1.01,0.13649280071258546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,2,power_law_1.01,0.5382207870483399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,2,power_law_1.01,0.16824959516525267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,2,power_law_1.01,0.8605631828308106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.01,0.19157760143280028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,2,power_law_1.01,0.9336895942687988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.01,0.24698240756988527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,2,power_law_1.01,1.4973440170288086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.01,0.2869312047958374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,2,power_law_1.01,2.1039615631103517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.01,0.38311679363250734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,2,power_law_1.01,2.6668607711791994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.01,0.4633984088897705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.01,0.6986176013946533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,2,power_law_1.01,3.808665466308594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.01,0.9419520378112793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,2,power_law_1.01,7.796825408935547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.01,1.3346943855285645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.01,1.4808768272399901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,2,power_law_1.01,2.0727039337158204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,2,power_law_1.01,2.9327680587768556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,2,power_law_1.01,6.222489547729492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,2,power_law_1.01,0.022355200350284578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,2,power_law_1.01,0.02364159971475601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,2,power_law_1.01,0.023897600173950196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,2,power_law_1.01,0.02510719895362854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,2,power_law_1.2,0.10463999509811402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,2,power_law_1.01,0.030822399258613586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,2,power_law_1.2,0.11281919479370117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,2,power_law_1.01,0.03079040050506592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,2,power_law_1.2,0.1518847942352295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,2,power_law_1.01,0.030048000812530517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,2,power_law_1.2,0.1768895983695984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,2,power_law_1.01,0.031455999612808226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,2,power_law_1.2,0.1738495945930481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,2,power_law_1.01,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,2,power_law_1.2,0.18213119506835937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,2,power_law_1.01,0.03272320032119751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,2,power_law_1.2,0.17077120542526245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,2,power_law_1.01,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,2,power_law_1.2,0.17319040298461913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,2,power_law_1.01,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,2,power_law_1.2,0.1712000012397766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,2,power_law_1.01,0.07336959838867188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,2,power_law_1.2,0.17555199861526488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,2,power_law_1.01,0.07822080254554749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,2,power_law_1.2,0.1678272008895874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,2,power_law_1.01,0.062054401636123656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,2,power_law_1.2,0.1980031967163086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,2,power_law_1.01,0.05334399938583374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,2,power_law_1.2,0.1832703948020935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,2,power_law_1.01,0.05839999914169312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,2,power_law_1.2,0.18725759983062745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,2,power_law_1.01,0.07294080257415772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,2,power_law_1.2,0.20544641017913817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.01,0.08894079923629761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,2,power_law_1.2,0.21690239906311035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.01,0.12119040489196778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,2,power_law_1.2,0.21643519401550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.01,0.14638079404830934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,2,power_law_1.2,0.296230411529541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.01,0.2030719995498657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,2,power_law_1.2,0.3410239934921265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.01,0.2571392059326172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,2,power_law_1.2,0.4560192108154297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.01,0.3669248104095459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,2,power_law_1.2,0.5197760105133057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.01,0.4738880157470703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,2,power_law_1.2,0.7099455833435059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.01,0.691648006439209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,2,power_law_1.2,0.07457280158996582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,2,power_law_1.2,0.8400959968566895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.01,0.9126848220825196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,2,power_law_1.2,0.07654399871826172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,2,power_law_1.2,1.4656255722045899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,2,power_law_1.01,1.1261247634887694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,2,power_law_1.2,0.07431039810180665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,2,power_law_1.2,1.6879039764404298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,2,power_law_1.01,1.782681655883789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,2,power_law_1.2,0.11568000316619872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,2,power_law_1.2,2.5053504943847655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,2,power_law_1.2,0.11059839725494384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,2,power_law_1.01,3.5093441009521484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,2,power_law_1.2,0.10901119709014892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,2,power_law_1.2,3.384998321533203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,2,power_law_1.2,0.11064319610595703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,2,power_law_1.2,3.9398208618164063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,2,power_law_1.2,0.07657600045204163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,2,power_law_1.2,0.11119359731674194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,2,power_law_1.2,0.08173440098762512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,2,power_law_1.2,6.263443374633789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,2,power_law_1.2,0.10718079805374145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,2,power_law_1.2,0.07597439885139465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,2,power_law_1.2,0.11037440299987793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,2,power_law_1.2,13.702841186523438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,2,power_law_1.2,0.09944959878921508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,2,power_law_1.2,0.10889600515365601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,2,power_law_1.2,0.10833280086517334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,2,power_law_1.2,0.118777596950531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,2,power_law_1.2,0.11468160152435303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,2,power_law_1.2,0.1219648003578186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,2,power_law_1.2,0.10799360275268555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,2,power_law_1.2,0.12360320091247559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,2,power_law_1.2,0.10799360275268555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,2,power_law_1.2,0.13498239517211913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,2,power_law_1.2,0.10599679946899414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,2,power_law_1.2,0.14009599685668944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,2,power_law_1.2,0.10428160429000854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,2,power_law_1.2,0.15497599840164183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,2,power_law_1.2,0.10327680110931396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,2,power_law_1.2,0.19309439659118652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,2,power_law_1.2,0.11297279596328735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,2,power_law_1.2,0.23001599311828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,2,power_law_1.2,0.11319040060043335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,2,power_law_1.2,0.2889280080795288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,2,power_law_1.2,0.1133504033088684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,2,power_law_1.2,0.3394176006317139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,2,power_law_1.2,0.13627519607543945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,2,power_law_1.2,0.45151357650756835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,2,power_law_1.2,0.13569920063018798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,2,power_law_1.2,0.5973504066467286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,2,power_law_1.2,0.152128005027771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,2,power_law_1.2,0.9011263847351074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,2,power_law_1.2,0.16401920318603516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.2,0.19043200016021727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,2,power_law_1.2,1.1568960189819335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.2,0.21403520107269286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,2,power_law_1.2,1.725369644165039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.2,0.26855039596557617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,2,power_law_1.2,2.188172721862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.2,0.3909440040588379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,2,power_law_1.2,2.4452159881591795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.2,0.5114367961883545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,2,power_law_1.2,4.392985534667969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.2,0.6420224189758301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.2,0.889305591583252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,2,power_law_1.2,7.794464111328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.2,1.3923263549804688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.2,1.6352384567260743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,2,power_law_1.2,2.238105583190918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,2,power_law_1.2,3.783718490600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,2,power_law_1.2,6.5048065185546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,2,power_law_1.2,0.02263679951429367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,2,power_law_1.2,0.022579200565814972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,2,power_law_1.2,0.02380799949169159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,2,power_law_1.2,0.024825599789619446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,2,power_law_1.2,0.030707201361656188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,1,balanced,0.08024533092975616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,2,power_law_1.2,0.028998398780822755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,1,balanced,0.10018666585286458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,2,power_law_1.2,0.0296640008687973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,1,balanced,0.1583146651585897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,2,power_law_1.2,0.03146879971027374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,1,balanced,0.1618826687335968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,2,power_law_1.2,0.03580799996852875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,1,balanced,0.16792533795038858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,2,power_law_1.2,0.03214719891548157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,1,balanced,0.16037333011627197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,2,power_law_1.2,0.035795199871063235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,1,balanced,0.18068800369898477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,2,power_law_1.2,0.07359359860420227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,1,balanced,0.1513759990533193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,2,power_law_1.2,0.07192959785461425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,1,balanced,0.16018666823705038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,2,power_law_1.2,0.07736960053443909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,1,balanced,0.16268799702326456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,2,power_law_1.2,0.06152960062026978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,1,balanced,0.16506666938463846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,2,power_law_1.2,0.05260800123214722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,1,balanced,0.14898666739463806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,2,power_law_1.2,0.05792639851570129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,1,balanced,0.1481706698735555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,2,power_law_1.2,0.07264639735221863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,1,balanced,0.14896000425020853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.2,0.08957440257072449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,1,balanced,0.1488053301970164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.2,0.12039040327072144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,1,balanced,0.14985600113868713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.2,0.14570239782333375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,1,balanced,0.15646933515866598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.2,0.20380799770355223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,1,balanced,0.22659200429916382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.2,0.25729920864105227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,1,balanced,0.21580799420674643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.2,0.3666048049926758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,1,balanced,0.2797013322512309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.2,0.4752255916595459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.2,0.6935872077941895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,1,balanced,0.32928532361984253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,1,balanced,0.0521919975678126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.2,0.9126463890075683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,1,balanced,0.07372800012429555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,1,balanced,0.46139732996622723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,2,power_law_1.2,1.1297087669372559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,1,balanced,0.0979146659374237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,1,balanced,0.10260799527168274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,1,balanced,0.5791840155919393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,2,power_law_1.2,1.7736000061035155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,1,balanced,0.1086133321126302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,2,power_law_1.2,3.5081600189208983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,1,balanced,0.1079306701819102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,1,balanced,0.8331092993418375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,1,balanced,0.11251200238863628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,1,balanced,0.10269332925478618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,1,balanced,1.0776267051696777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,1,balanced,0.11120532949765523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,1,balanced,0.10487467050552368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,1,balanced,0.0572320024172465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,1,balanced,0.09950400392214458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,1,balanced,0.0830506682395935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,1,balanced,0.10427199800809224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,1,balanced,1.6024479866027832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,1,balanced,0.10250666737556458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,1,balanced,0.10215466221173604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,1,balanced,0.10739200313886006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,1,balanced,0.10188266634941101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,1,balanced,0.10472533106803894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,1,balanced,0.10846933722496033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,1,balanced,0.11829333504041036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,1,balanced,0.1088213324546814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,1,balanced,0.1058240036169688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,1,balanced,2.1104586919148765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,1,balanced,0.11528533697128296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,1,balanced,0.11102933684984843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,1,balanced,0.10150399804115295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,1,balanced,0.15616533160209656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,1,balanced,0.1046506663163503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,1,balanced,0.16064000129699707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,1,balanced,0.11214933792750041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,1,balanced,0.20239466428756714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,1,balanced,2.6310826937357583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,1,balanced,0.10656533638636272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,1,balanced,0.1002453366915385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,1,balanced,0.2437386711438497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,1,balanced,0.10080533226331075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,1,balanced,0.33537598450978595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,1,balanced,0.10753066341082256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,1,balanced,0.09594133496284485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,1,balanced,0.42842666308085126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,1,balanced,0.09859733780225118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,1,balanced,4.253189404805501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,1,balanced,0.15076800187428793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,1,balanced,0.6134080092112223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,1,balanced,0.1426400045553843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,1,balanced,0.17896533012390137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,1,balanced,0.7957066694895426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,1,balanced,0.21702933311462402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,1,balanced,0.29305599133173627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,1,balanced,1.1584320068359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,1,balanced,0.37542935212453205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,1,balanced,0.529312014579773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,1,balanced,8.499173482259115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,1,balanced,1.5395946502685547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,1,balanced,0.6863306363423666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,1,balanced,1.9099680582682292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,1,balanced,0.9953333536783854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,1,balanced,1.3190399805704753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,1,balanced,3.0386826197306314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,1,balanced,1.6282933553059895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,1,balanced,5.991984049479167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,1,balanced,2.5452319780985513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,1,balanced,0.022261333962281544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,1,balanced,0.02256533255179723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,1,balanced,0.02422933280467987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,1,balanced,0.024821333587169647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,1,balanced,0.026682667434215546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,1,balanced,0.0308693324526151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,1,balanced,0.02978666623433431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,1,balanced,0.032101333141326904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,1,balanced,0.031727999448776245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,1,balanced,0.03234666585922241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,1,balanced,5.063151995340983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,1,balanced,0.032933334509531655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,1,balanced,0.035274667044480644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,1,balanced,0.037104000647862755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,1,balanced,0.03821333249409994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,1,balanced,0.07627200086911519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,1,balanced,0.07664533456166585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,1,balanced,0.08003200093905131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,1,balanced,0.058277333776156105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,1,balanced,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,1,balanced,0.08320533235867818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,1,balanced,0.10120532910029094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,1,balanced,0.13512000441551208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,1,balanced,0.16192000110944113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,1,balanced,0.2337013284365336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,1,balanced,0.2935413320859273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,1,balanced,0.42421865463256836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,1,balanced,0.5500213305155436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,1,power_law_1.01,0.07134079933166504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,1,power_law_1.01,0.09587839841842652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,1,power_law_1.01,0.11350400447845459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,1,power_law_1.01,0.15415680408477783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,1,power_law_1.01,0.15381759405136108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,1,power_law_1.01,0.16108800172805787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,1,balanced,0.6807413101196289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,1,power_law_1.01,0.16318080425262452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,1,power_law_1.01,0.16483839750289916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,1,power_law_1.01,0.15903359651565552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,1,power_law_1.01,0.1549183964729309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,1,power_law_1.01,0.16181119680404663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,1,power_law_1.01,0.1561344027519226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,1,balanced,1.0668106873830159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,1,power_law_1.01,0.1673472046852112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,1,power_law_1.01,0.173471999168396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,1,power_law_1.01,0.19059840440750123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,1,power_law_1.01,0.19404159784317015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,1,power_law_1.01,0.21255040168762207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,1,power_law_1.01,0.24196479320526124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,1,power_law_1.01,0.2729088068008423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,1,power_law_1.01,0.34133119583129884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,1,power_law_1.01,0.40497279167175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,1,power_law_1.01,0.5271999835968018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,1,power_law_1.01,0.6716095924377441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,1,power_law_1.01,0.05777919888496399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,1,power_law_1.01,0.9073087692260742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,1,power_law_1.01,0.06388480067253113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,1,power_law_1.01,1.1650943756103516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,1,balanced,2.086293379465739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,1,power_law_1.01,0.0753216028213501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,1,power_law_1.01,1.691481590270996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,1,power_law_1.01,0.09997439980506898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,1,power_law_1.01,0.10261119604110717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,1,power_law_1.01,2.1712127685546876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,1,power_law_1.01,0.10354559421539307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,1,power_law_1.01,2.727840042114258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,1,power_law_1.01,0.10303360223770142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,1,power_law_1.01,4.217478561401367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,1,power_law_1.01,0.10120960474014282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,1,power_law_1.01,0.10094079971313477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,1,power_law_1.01,8.605427551269532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,1,power_law_1.01,0.10448000431060792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,1,power_law_1.01,0.10316159725189208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,1,power_law_1.01,0.10582400560379028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,1,power_law_1.01,0.10801279544830322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,1,power_law_1.01,0.11162879467010497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,1,power_law_1.01,0.1285823941230774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,1,power_law_1.01,0.058246397972106935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,1,power_law_1.01,0.1313599944114685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,1,power_law_1.01,0.07010560035705567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,1,power_law_1.01,0.14136960506439208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,1,power_law_1.01,0.07263360023498536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,1,power_law_1.01,0.16533119678497316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,1,power_law_1.01,0.10047359466552734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,1,power_law_1.01,0.18473600149154662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,1,power_law_1.01,0.10341759920120239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,1,power_law_1.01,0.24760959148406983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,1,power_law_1.01,0.11050879955291748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,1,power_law_1.01,0.27699201107025145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,1,power_law_1.01,0.10666879415512084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,1,power_law_1.01,0.3766655921936035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,1,power_law_1.01,0.10462080240249634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,1,power_law_1.01,0.47124481201171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,1,power_law_1.01,0.10207359790802002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,1,power_law_1.01,0.6551936149597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,1,power_law_1.01,0.11030399799346924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,1,power_law_1.01,0.8309823989868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,1,power_law_1.01,0.10490880012512208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,1,power_law_1.01,1.20251522064209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,1,power_law_1.01,0.10485759973526002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,1,power_law_1.01,1.5633536338806153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,1,power_law_1.01,0.11292799711227416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,1,power_law_1.01,1.9399999618530273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,1,power_law_1.01,0.11278719902038574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,1,power_law_1.01,0.12257920503616333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,1,power_law_1.01,3.0688575744628905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,1,power_law_1.01,0.12947200536727904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,1,power_law_1.01,6.023628616333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,1,power_law_1.01,0.13575040102005004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,1,power_law_1.01,0.15847040414810182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.01,0.18495999574661254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.01,0.22410879135131836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.01,0.2622335910797119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.01,0.33384320735931394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.01,0.4134655952453613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.01,0.5707647800445557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.01,0.7330304145812988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.01,1.019040012359619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.01,1.3406271934509277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.01,1.6445375442504884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,1,power_law_1.01,2.604198455810547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,1,power_law_1.01,5.09191665649414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,1,power_law_1.01,0.020563200116157532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,1,power_law_1.01,0.021798400580883025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,1,power_law_1.01,0.02242559939622879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,1,power_law_1.01,0.023270399868488313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,1,power_law_1.01,0.02497279942035675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,1,power_law_1.01,0.02980479896068573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,1,power_law_1.01,0.02876800000667572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,1,power_law_1.01,0.02980479896068573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,1,power_law_1.2,0.07445120215415954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,1,power_law_1.01,0.03057279884815216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,1,power_law_1.2,0.09003520011901855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,1,power_law_1.01,0.030131199955940248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,1,power_law_1.2,0.08892160058021545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,1,power_law_1.01,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,1,power_law_1.2,0.15285120010375977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,1,power_law_1.01,0.034355199337005614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,1,power_law_1.2,0.16506880521774292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,1,power_law_1.01,0.03439359962940216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,1,power_law_1.2,0.15599360466003417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,1,power_law_1.01,0.03651840090751648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,1,power_law_1.2,0.15841280221939086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,1,power_law_1.01,0.07344639897346497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,1,power_law_1.2,0.16010240316390992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,1,power_law_1.01,0.07522559762001038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,1,power_law_1.2,0.16090240478515624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,1,power_law_1.01,0.07920640110969543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,1,power_law_1.2,0.1591871976852417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,1,power_law_1.01,0.056441599130630495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,1,power_law_1.2,0.160806405544281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.01,0.06123520135879516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,1,power_law_1.2,0.16741759777069093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.01,0.08273919820785522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,1,power_law_1.2,0.16151039600372313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.01,0.10126719474792481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,1,power_law_1.2,0.17772799730300903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.01,0.13256959915161132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,1,power_law_1.2,0.19423999786376953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.01,0.16075520515441893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,1,power_law_1.2,0.18807040452957152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.01,0.23148159980773925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,1,power_law_1.2,0.21015679836273193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.01,0.29482879638671877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,1,power_law_1.2,0.23601920604705812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.01,0.4230527877807617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,1,power_law_1.2,0.2803328037261963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.01,0.5492991924285888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,1,power_law_1.2,0.34225919246673586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.01,0.6803520202636719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,1,power_law_1.2,0.4090240001678467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,1,power_law_1.01,1.0667391777038575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,1,power_law_1.2,0.5359360218048096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,1,power_law_1.2,0.06054400205612183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,1,power_law_1.01,2.084524726867676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,1,power_law_1.2,0.6683263778686523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,1,power_law_1.2,0.06212480068206787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,1,power_law_1.2,0.9165632247924804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,1,power_law_1.2,0.06816639900207519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,1,power_law_1.2,1.1338751792907715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,1,power_law_1.2,0.10474239587783814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,1,power_law_1.2,1.6641536712646485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,1,power_law_1.2,0.10303360223770142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,1,power_law_1.2,0.10325759649276733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,1,power_law_1.2,2.2016639709472656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,1,power_law_1.2,0.10252799987792968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,1,power_law_1.2,2.6954816818237304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,1,power_law_1.2,0.10180480480194092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,1,power_law_1.2,4.278163146972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,1,power_law_1.2,0.10229120254516602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,1,power_law_1.2,0.10462720394134521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,1,power_law_1.2,8.548921966552735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,1,power_law_1.2,0.10106240510940552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,1,power_law_1.2,0.10696320533752442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,1,power_law_1.2,0.10841599702835084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,1,power_law_1.2,0.122489595413208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,1,power_law_1.2,0.12527999877929688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,1,power_law_1.2,0.13015040159225463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,1,power_law_1.2,0.056806397438049314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,1,power_law_1.2,0.14616960287094116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,1,power_law_1.2,0.06812800168991089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,1,power_law_1.2,0.16778240203857422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,1,power_law_1.2,0.07360000014305115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,1,power_law_1.2,0.18838399648666382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,1,power_law_1.2,0.1146880030632019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,1,power_law_1.2,0.22812159061431886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,1,power_law_1.2,0.10512000322341919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,1,power_law_1.2,0.279200005531311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,1,power_law_1.2,0.10303360223770142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,1,power_law_1.2,0.37733120918273927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,1,power_law_1.2,0.10472960472106933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,1,power_law_1.2,0.45981440544128416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,1,power_law_1.2,0.10311679840087891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,1,power_law_1.2,0.6539008140563964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,1,power_law_1.2,0.10077439546585083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,1,power_law_1.2,0.8365119934082031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,1,power_law_1.2,0.10172799825668336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,1,power_law_1.2,1.1944895744323731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,1,power_law_1.2,0.10160640478134156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,1,power_law_1.2,1.5624768257141113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,1,power_law_1.2,0.11011840105056762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,1,power_law_1.2,1.9378944396972657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,1,power_law_1.2,0.1102720022201538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,1,power_law_1.2,3.0514368057250976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,1,power_law_1.2,0.11465599536895751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,1,power_law_1.2,6.025388717651367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,1,power_law_1.2,0.11909760236740112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,1,power_law_1.2,0.1301695942878723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,1,power_law_1.2,0.1335744023323059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,1,power_law_1.2,0.15431679487228395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.2,0.18094079494476317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.2,0.21797759532928468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.2,0.25306880474090576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.2,0.33218560218811033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.2,0.41228160858154295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.2,0.5641727924346924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.2,0.7016448020935059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.2,1.0273280143737793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.2,1.34268798828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.2,1.6324352264404296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,1,power_law_1.2,2.5733312606811523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,1,power_law_1.2,5.087654495239258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,1,power_law_1.2,0.02059520035982132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,1,power_law_1.2,0.02192640006542206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,1,power_law_1.2,0.022137600183486938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,1,power_law_1.2,0.023520000278949738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,1,power_law_1.2,0.024883200228214265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,1,power_law_1.2,0.029337599873542786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,1,power_law_1.2,0.028672000765800475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,1,power_law_1.2,0.029683199524879456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,1,power_law_1.2,0.029068800806999206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,1,power_law_1.2,0.03020800054073334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,1,power_law_1.2,0.03182719945907593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,1,power_law_1.2,0.03162879943847656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,1,power_law_1.2,0.03346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,1,power_law_1.2,0.036582401394844054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,1,power_law_1.2,0.07352319955825806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,1,power_law_1.2,0.07345920205116271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,1,power_law_1.2,0.07711359858512878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,1,power_law_1.2,0.05589759945869446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.2,0.06047359704971313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.2,0.08041599988937378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.2,0.09867519736289979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.2,0.13374719619750977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.2,0.15968639850616456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.2,0.23020799160003663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.2,0.2918976068496704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.2,0.42177281379699705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.2,0.5520832061767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.2,0.6798399925231934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,1,power_law_1.2,1.0640447616577149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,1,power_law_1.2,2.084192085266113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,8,balanced,0.08773866295814514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,8,balanced,0.10054933031400044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,8,balanced,0.12296000123023987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,8,balanced,0.18847467501958212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,8,balanced,0.33056533336639404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,8,balanced,0.538703997929891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,8,balanced,0.550378680229187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,8,balanced,0.5558773279190063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,8,balanced,0.5406560103098551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,8,balanced,0.573194662729899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,8,balanced,0.5394560098648071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,8,balanced,0.5668373505274454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,8,balanced,0.5993706782658895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,8,balanced,0.5333066781361898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,8,balanced,0.6054933468500773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,8,balanced,0.5724853277206421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,8,balanced,0.5297919909159342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,8,balanced,0.543498675028483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,8,balanced,0.5549920002619425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,8,balanced,0.5540586709976196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,8,balanced,0.06376533210277557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,8,balanced,0.07904000083605449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,8,balanced,0.090421328941981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,8,balanced,0.5599626700083414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,8,balanced,0.13402133186658224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,8,balanced,0.20482132832209268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,8,balanced,0.3454666535059611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,8,balanced,0.32862399021784466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,8,balanced,0.5831466515858968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,8,balanced,0.3234986662864685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,8,balanced,0.31613866488138836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,8,balanced,0.3222773273785909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,8,balanced,0.3110613425572713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,8,balanced,0.617413322130839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,8,balanced,0.316538671652476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,8,balanced,0.336736003557841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,8,balanced,0.06486399968465169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,8,balanced,0.3083946704864502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,8,balanced,0.07408533493677776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,8,balanced,0.3282986680666606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,8,balanced,0.093231995900472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,8,balanced,0.14708800117174783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,8,balanced,0.34059735139211017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,8,balanced,0.20150933663050333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,8,balanced,0.3274986743927002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,8,balanced,0.9400533040364584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,8,balanced,0.3673866589864095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,8,balanced,0.32840534051259357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,8,balanced,0.3397386471430461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,8,balanced,0.33210132519404095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,8,balanced,0.31749866406122845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,8,balanced,0.34959999720255536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,8,balanced,0.3197866678237915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,8,balanced,0.30805333455403644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,8,balanced,0.37007999420166016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,8,balanced,0.2977280020713806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,8,balanced,0.3269600073496501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,8,balanced,0.9210240046183268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,8,balanced,0.31933865944544476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,8,balanced,0.3773760000864665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,8,balanced,0.3110239903132121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,8,balanced,0.34386666615804035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,8,balanced,0.31330666939417523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,8,balanced,0.42268800735473633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,8,balanced,0.3561386664708455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,8,balanced,0.3257066607475281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,8,balanced,0.46905601024627686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,8,balanced,0.3399146795272827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,8,balanced,1.2698720296223958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,8,balanced,0.33376534779866535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,8,balanced,0.7123520374298096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,8,balanced,0.35412800312042236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,8,balanced,0.3842453161875407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,8,balanced,0.7452320257822672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,8,balanced,0.6336959997812907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,8,balanced,1.6084373792012532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,8,balanced,1.0415039857228596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,8,balanced,0.6383839845657349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,8,balanced,0.8898613452911377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,8,balanced,1.9723893801371257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,8,balanced,0.041984001795450844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,8,balanced,1.3254986604054768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,8,balanced,0.044938668608665466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,8,balanced,0.0517493337392807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,8,balanced,0.06159999966621399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,8,balanced,0.08796800176302592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,8,balanced,0.14165332913398743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,8,balanced,0.14264532923698425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,8,balanced,0.1461120049158732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,8,balanced,0.16562133034070334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,8,balanced,1.1481599807739258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,8,balanced,0.16145599881807962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,8,balanced,0.1660480002562205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,8,balanced,0.24259734153747559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,8,balanced,0.22664000590642294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,8,balanced,0.23146667083104452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,8,balanced,0.31405866146087646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,8,balanced,0.3192906578381856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,8,balanced,1.6324693361918132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,8,balanced,0.34988800684611004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,8,balanced,0.31523199876149494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,8,balanced,0.33368531862894696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,8,balanced,1.403439998626709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,8,balanced,0.3716213305791219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,8,balanced,3.071551958719889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,8,balanced,0.4131840070088704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,8,balanced,0.4724693298339844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,8,balanced,0.5476266543070475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,8,balanced,2.531770706176758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,8,balanced,0.6681546370188395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,8,balanced,2.169007937113444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,8,balanced,0.8274827003479004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,8,balanced,1.3162026405334473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,8,balanced,6.084197362263997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,8,balanced,1.5787893931070964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,8,balanced,4.2492319742838545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,8,balanced,5.006661415100098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,8,balanced,2.0687146186828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,8,balanced,3.072304089864095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,8,power_law_1.01,0.19214080572128295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,8,power_law_1.01,0.3166208028793335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,8,power_law_1.01,0.17532800436019896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,8,power_law_1.01,0.18587520122528076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,8,power_law_1.01,0.26149759292602537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,8,power_law_1.01,0.3647871971130371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,8,power_law_1.01,0.43077759742736815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,8,power_law_1.01,0.44785919189453127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,8,balanced,6.065413157145183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,8,power_law_1.01,0.4731135845184326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,8,power_law_1.01,0.45808000564575196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,8,power_law_1.01,0.46036481857299805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,8,power_law_1.01,0.5118591785430908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,8,power_law_1.01,0.14054399728775024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,8,power_law_1.01,0.48638081550598145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,8,power_law_1.01,0.21802880764007568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,8,power_law_1.01,0.504960012435913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,8,power_law_1.01,0.13315199613571166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,8,power_law_1.01,0.13980159759521485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,8,power_law_1.01,0.535539197921753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,8,power_law_1.01,0.21086080074310304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,8,power_law_1.01,0.12743040323257446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,8,power_law_1.01,0.504915189743042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,8,power_law_1.01,0.11201280355453491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,8,power_law_1.01,0.18496639728546144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,8,power_law_1.01,0.5122496128082276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,8,power_law_1.01,0.11938560009002686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,8,power_law_1.01,0.2157439947128296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,8,power_law_1.01,0.541209602355957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,8,power_law_1.01,0.16435840129852294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,8,power_law_1.01,0.27820160388946535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,8,power_law_1.01,0.5907519817352295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,8,power_law_1.01,0.20158720016479492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,8,power_law_1.01,0.2706815958023071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,8,power_law_1.01,0.6661888122558594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,8,power_law_1.01,0.26168320178985593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,8,power_law_1.01,0.2805952072143555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,8,power_law_1.01,0.7019455909729004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,8,power_law_1.01,0.28312320709228517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,8,power_law_1.01,0.2813760042190552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,8,power_law_1.01,0.8822527885437011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,8,power_law_1.01,0.2737279891967773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,8,power_law_1.01,0.281497597694397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,8,power_law_1.01,0.2997567892074585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,8,power_law_1.01,0.2932800054550171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,8,power_law_1.01,0.9857536315917969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,8,power_law_1.01,0.29141759872436523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,8,power_law_1.01,1.2674943923950195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,8,power_law_1.01,0.30033919811248777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,8,power_law_1.01,1.5473471641540528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,8,power_law_1.01,0.3018815994262695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,8,power_law_1.01,2.118560028076172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,8,power_law_1.01,0.3201280117034912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,8,power_law_1.01,2.7603456497192385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,8,power_law_1.01,0.3153280019760132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,8,power_law_1.01,3.1307903289794923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,8,power_law_1.01,0.34741759300231934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.01,0.364851188659668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,8,power_law_1.01,4.596480178833008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.01,0.40945281982421877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,8,power_law_1.01,8.94427490234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.01,0.45292158126831056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.01,0.5370495796203614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.01,0.6195328235626221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.01,0.804319953918457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.01,0.9627455711364746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.01,1.3447232246398926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.01,1.7253887176513671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,8,power_law_1.01,1.929484748840332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,8,power_law_1.01,2.965318489074707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,8,power_law_1.01,5.486963272094727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,8,power_law_1.01,0.047628799080848695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,8,power_law_1.01,0.07573760151863099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,8,power_law_1.01,0.05317760109901428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,8,power_law_1.01,0.06225280165672302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,8,power_law_1.01,0.07367680072784424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,8,power_law_1.01,0.08677120208740234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,8,power_law_1.01,0.11189759969711303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,8,power_law_1.01,0.10606080293655396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,8,power_law_1.01,0.12557439804077147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,8,power_law_1.01,0.12617599964141846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,8,power_law_1.01,0.12405120134353638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,8,power_law_1.01,0.1663807988166809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,8,power_law_1.01,0.16095999479293824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,8,power_law_1.01,0.1647104024887085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,8,power_law_1.01,0.2455296039581299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,8,power_law_1.01,0.2529792070388794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,8,power_law_1.01,0.268121600151062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,8,power_law_1.01,0.2573823928833008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.01,0.24471681118011473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.01,0.24728960990905763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,8,power_law_1.2,0.18897279500961303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.01,0.27253758907318115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,8,power_law_1.2,0.2884160041809082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.01,0.3126336097717285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,8,power_law_1.2,0.1211840033531189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.01,0.37535359859466555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,8,power_law_1.2,0.21006081104278565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,8,power_law_1.2,0.2598655939102173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.01,0.4590591907501221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,8,power_law_1.2,0.33023359775543215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.01,0.5963456153869628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,8,power_law_1.2,0.45395841598510744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.01,0.8220992088317871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,8,power_law_1.2,0.4437248229980469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.01,1.031315231323242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,8,power_law_1.2,0.45452160835266114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,8,power_law_1.01,1.3406271934509277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,8,power_law_1.2,0.45326080322265627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,8,power_law_1.01,1.9169599533081054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,8,power_law_1.2,0.4704768180847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,8,power_law_1.2,0.48302721977233887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,8,power_law_1.01,3.881849670410156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,8,power_law_1.2,0.5091455936431885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,8,power_law_1.01,0.28928000926971437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,8,power_law_1.2,0.5338431835174561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,8,power_law_1.01,0.3091455936431885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,8,power_law_1.2,0.5135295867919922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,8,power_law_1.2,0.12691839933395385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,8,power_law_1.01,0.29127039909362795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,8,power_law_1.2,0.5208576202392579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,8,power_law_1.01,0.3038719892501831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,8,power_law_1.2,0.19356160163879393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,8,power_law_1.2,0.5442048072814941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,8,power_law_1.01,0.31626238822937014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,8,power_law_1.2,0.08809599876403809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,8,power_law_1.01,0.32666239738464353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,8,power_law_1.2,0.583135986328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,8,power_law_1.2,0.11811840534210205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,8,power_law_1.01,0.3266239881515503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,8,power_law_1.2,0.6424448013305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,8,power_law_1.2,0.1555456042289734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,8,power_law_1.01,0.3385855913162231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,8,power_law_1.2,0.6921535968780518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,8,power_law_1.2,0.20300159454345704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,8,power_law_1.01,0.371891188621521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,8,power_law_1.2,0.2593535900115967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,8,power_law_1.2,0.8001664161682129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,8,power_law_1.2,0.2529792070388794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,8,power_law_1.01,0.41431679725646975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,8,power_law_1.2,0.9467904090881347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,8,power_law_1.2,0.2872704029083252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,8,power_law_1.01,0.48578557968139646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,8,power_law_1.2,1.1344703674316405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,8,power_law_1.01,0.6153408050537109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,8,power_law_1.2,0.27981441020965575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,8,power_law_1.2,1.3633407592773437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,8,power_law_1.01,0.6953792095184326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,8,power_law_1.2,0.27714560031890867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,8,power_law_1.2,1.7362560272216796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,8,power_law_1.2,0.29300479888916015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,8,power_law_1.01,0.8890751838684082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,8,power_law_1.2,2.3575103759765623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,8,power_law_1.2,0.3099391937255859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,8,power_law_1.01,1.1051456451416015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,8,power_law_1.2,3.1766336441040037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,8,power_law_1.2,0.31982719898223877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,8,power_law_1.01,1.5351167678833009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,8,power_law_1.2,3.509395217895508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,8,power_law_1.2,0.3035327911376953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,8,power_law_1.01,1.9202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,8,power_law_1.2,0.3261375904083252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,8,power_law_1.2,6.2504833221435545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,8,power_law_1.01,2.243539237976074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,8,power_law_1.2,0.3538111925125122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,8,power_law_1.01,3.524262237548828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,8,power_law_1.2,12.767123413085937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,8,power_law_1.2,0.36761600971221925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,8,power_law_1.2,0.40204801559448244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,8,power_law_1.01,6.68243179321289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,8,power_law_1.2,0.4359936237335205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,8,power_law_1.2,0.5278592109680176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,8,power_law_1.2,0.644704008102417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,8,power_law_1.2,0.7851071834564209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,8,power_law_1.2,0.941708755493164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,8,power_law_1.2,1.2135295867919922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,8,power_law_1.2,1.7136831283569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,8,power_law_1.2,2.144486427307129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,8,power_law_1.2,2.7421951293945312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,8,power_law_1.2,4.459993743896485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,8,power_law_1.2,8.544185638427734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,8,power_law_1.2,0.1389631986618042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,8,power_law_1.2,0.18476799726486207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,8,power_law_1.2,0.10043519735336304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,8,power_law_1.2,0.14364800453186036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,8,power_law_1.2,0.18702080249786376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,8,power_law_1.2,0.19029120206832886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,8,power_law_1.2,0.2484800100326538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,8,power_law_1.2,0.2920192003250122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,8,power_law_1.2,0.28384640216827395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,8,power_law_1.2,0.2817919969558716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,8,power_law_1.2,0.28487040996551516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,8,power_law_1.2,0.28949120044708254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,8,power_law_1.2,0.29954559803009034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,8,power_law_1.2,0.3097856044769287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,8,power_law_1.2,0.3187903881072998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,8,power_law_1.2,0.3188479900360107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,8,power_law_1.2,0.3408447980880737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,8,power_law_1.2,0.3462080001831055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.2,0.37443840503692627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.2,0.420684814453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,4,balanced,0.07675200204054515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.2,0.48037118911743165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,4,balanced,0.09353066484133403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.2,0.6288191795349121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,4,balanced,0.12099732955296834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,4,balanced,0.18412800629933676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.2,0.6778687953948974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,4,balanced,0.3156213363011678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.2,0.9343680381774903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,4,balanced,0.533893346786499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,4,balanced,0.5258400042851766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.2,1.0772159576416016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,8,power_law_1.2,0.04737919867038727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,4,balanced,0.5228960116704305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,8,power_law_1.2,0.07399680018424988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.2,1.5217408180236816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,4,balanced,0.525002678235372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,8,power_law_1.2,0.04619520008563995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,4,balanced,0.5583893458048502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.2,1.9188735961914063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,4,balanced,0.5085119803746542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,8,power_law_1.2,0.059622400999069215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,8,power_law_1.2,2.2109376907348635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,4,balanced,0.524778683980306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,8,power_law_1.2,0.06776959896087646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,4,balanced,0.5528639952341715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,8,power_law_1.2,3.346169662475586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,8,power_law_1.2,0.08186240196228027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,4,balanced,0.5314559936523438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,8,power_law_1.2,0.1087231993675232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,8,power_law_1.2,7.699858856201172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,4,balanced,0.5535733302434286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,8,power_law_1.2,0.11088000535964966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,4,balanced,0.5553760131200155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,8,power_law_1.2,0.12349439859390259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,4,balanced,0.5250933170318604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,8,power_law_1.2,0.12984319925308227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,4,balanced,0.566490650177002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,8,power_law_1.2,0.12828799486160278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,8,power_law_1.2,0.17370879650115967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,4,balanced,0.5419679880142212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,4,balanced,0.06488533318042755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,8,power_law_1.2,0.1748095989227295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,4,balanced,0.07202666501204173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,4,balanced,0.5684106747309366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,8,power_law_1.2,0.14954240322113038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,4,balanced,0.08868799606959026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,4,balanced,0.12705600261688232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,8,power_law_1.2,0.24865279197692872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,4,balanced,0.20524267355600992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,4,balanced,0.5819146633148193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,8,power_law_1.2,0.2536639928817749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,4,balanced,0.3031040032704671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,8,power_law_1.2,0.2737152099609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,4,balanced,0.30959467093149823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,8,power_law_1.2,0.2587007999420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,4,balanced,0.30899200836817425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,4,balanced,0.623637318611145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,4,balanced,0.29789867003758747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.2,0.2678015947341919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,4,balanced,0.30184000730514526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.2,0.26360321044921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,4,balanced,0.30471465984980267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,4,balanced,0.3017866611480713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.2,0.2895103931427002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,4,balanced,0.6715146700541178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,4,balanced,0.3283360004425049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.2,0.3448767900466919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,4,balanced,0.3144320050875346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.2,0.3968384027481079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,4,balanced,0.3155360023180644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.2,0.5466879844665528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,4,balanced,0.33218133449554443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.2,0.6124800205230713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,4,balanced,0.32996799548467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,4,balanced,1.0001227060953777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.2,0.8581376075744629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,4,balanced,0.34827200571695965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.2,1.1077376365661622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,4,balanced,0.3540480136871338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,8,power_law_1.2,1.5001536369323731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,4,balanced,0.3825920025507609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,8,power_law_1.2,2.284422492980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,4,balanced,1.0162399609883626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,8,power_law_1.2,4.584422302246094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,4,balanced,0.4069066842397054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,4,balanced,0.46825067202250165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,4,balanced,0.5325066645940145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,4,balanced,1.413306713104248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,4,balanced,0.7751093705495199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,4,balanced,0.8718187014261881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,4,balanced,1.7796053886413574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,4,balanced,1.2196426391601562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,4,balanced,1.559050718943278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,4,balanced,2.1737546920776367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,4,balanced,1.9222985903422039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,4,balanced,3.4401706059773765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,4,balanced,0.06885333359241486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,4,balanced,0.08213333288828532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,4,balanced,0.09558399518330891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,4,balanced,0.15173332889874777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,4,balanced,0.2183199922243754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,4,balanced,0.3589973449707031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,4,balanced,3.032719930013021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,4,balanced,0.3351680040359497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,4,balanced,0.33000532786051434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,4,balanced,0.33060266574223834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,4,balanced,0.3167840043703715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,4,balanced,0.32264532645543414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,4,balanced,0.31275733311971027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,4,balanced,0.3165760040283203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,4,balanced,0.3041226665178935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,4,balanced,0.3190986712773641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,4,balanced,0.34437334537506104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,4,balanced,0.31755199035008747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,4,balanced,0.3378880023956299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,4,balanced,0.33984001477559406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,4,balanced,0.3614293336868286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,4,balanced,0.370527982711792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,4,balanced,0.02773333340883255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,4,balanced,0.41606398423512775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,4,balanced,0.03073599934577942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,4,balanced,0.04675200084845225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,4,balanced,0.05935466786225637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,4,balanced,0.08683199683825175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,4,balanced,0.45846398671468097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,4,balanced,0.1381760040918986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,4,balanced,0.14264532923698425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,4,balanced,0.15054933230082193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,4,balanced,0.14656533797581991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,4,balanced,0.14857600132624307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,4,balanced,0.15467199683189392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,4,balanced,0.7244640191396078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,4,balanced,0.16671466827392578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,4,balanced,0.16755733887354532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,4,balanced,6.858090718587239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,4,balanced,0.17377599080403647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,4,balanced,0.2337119976679484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,4,balanced,0.23437867561976114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,4,balanced,5.9855092366536455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,4,balanced,0.236026664574941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,4,balanced,0.7853546937306722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,4,balanced,0.34411199887593585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,4,balanced,0.3857119878133138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,4,balanced,0.34840532143910724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,4,balanced,0.3708266814549764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,4,balanced,1.099845329920451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,4,balanced,0.41417598724365234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,4,balanced,0.45602667331695557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,4,balanced,0.5333600044250488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,4,balanced,1.4337600072224934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,4,balanced,0.6080960035324097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,4,balanced,1.7617653210957844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,4,balanced,0.9860479831695557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,4,balanced,1.1566452980041504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,4,balanced,2.728581428527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,4,balanced,1.5394879976908367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,4,balanced,2.2381866772969565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,4,power_law_1.01,0.1273344039916992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,4,power_law_1.01,0.19823999404907228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,4,power_law_1.01,0.11436799764633179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,4,power_law_1.01,0.16147199869155884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,4,power_law_1.01,0.08958719968795777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,4,power_law_1.01,0.23610239028930663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,4,power_law_1.01,0.12680959701538086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,4,power_law_1.01,0.3088831901550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,4,power_law_1.01,0.09012479782104492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,4,power_law_1.01,0.4020864009857178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,4,power_law_1.01,0.11644799709320068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,4,power_law_1.01,0.43099517822265626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,4,power_law_1.01,0.4366144180297852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,4,power_law_1.01,0.15515520572662353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,4,power_law_1.01,0.4397183895111084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,4,power_law_1.01,0.18551679849624633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,4,power_law_1.01,0.4679743766784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,4,power_law_1.01,0.23452799320220946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,4,power_law_1.01,0.4688511848449707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,4,power_law_1.01,0.24693760871887208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,4,power_law_1.01,0.46639361381530764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,4,power_law_1.01,0.25793919563293455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,4,balanced,5.391066869099935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,4,power_law_1.01,0.49848318099975586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,4,power_law_1.01,0.2678335905075073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,4,power_law_1.01,0.4886911869049072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,4,power_law_1.01,0.28060159683227537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,4,power_law_1.01,0.5010560035705567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,4,power_law_1.01,0.2825151920318604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,4,power_law_1.01,0.49466238021850584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,4,power_law_1.01,0.28142719268798827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,4,power_law_1.01,0.5405312061309815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,4,power_law_1.01,0.30078079700469973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,4,power_law_1.01,0.5621632099151611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,4,power_law_1.01,0.29351038932800294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,4,power_law_1.01,0.631167984008789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,4,power_law_1.01,0.30512640476226804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,4,power_law_1.01,0.7071104049682617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,4,power_law_1.01,0.31271040439605713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,4,power_law_1.01,0.34900479316711425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,4,power_law_1.01,0.8608960151672364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,4,power_law_1.01,0.3726464033126831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,4,power_law_1.01,0.9662015914916993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,4,power_law_1.01,0.4215231895446777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,4,power_law_1.01,1.2577664375305175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,4,power_law_1.01,0.4792960166931152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,4,power_law_1.01,1.4606528282165527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,4,power_law_1.01,0.5832320213317871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,4,power_law_1.01,2.0427776336669923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,4,power_law_1.01,0.6666624069213867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,4,power_law_1.01,2.413279914855957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,4,balanced,4.430565198262532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,4,power_law_1.01,0.8912320137023926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,4,power_law_1.01,3.0622016906738283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,4,power_law_1.01,1.1045184135437012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,4,power_law_1.01,4.599391937255859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,4,power_law_1.01,1.5884223937988282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,4,power_law_1.01,1.9718847274780273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,4,power_law_1.01,8.964736175537109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,4,power_law_1.01,2.398451232910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,4,power_law_1.01,3.7377662658691406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,4,power_law_1.01,7.560594940185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,4,power_law_1.01,0.09758080244064331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,4,power_law_1.01,0.1369920015335083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,4,power_law_1.01,0.09629439711570739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,4,power_law_1.01,0.11059839725494384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,4,power_law_1.01,0.1647487998008728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,4,power_law_1.01,0.22483201026916505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,4,power_law_1.01,0.2579008102416992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,4,power_law_1.01,0.27742719650268555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,4,power_law_1.01,0.2710400104522705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,4,power_law_1.01,0.28958721160888673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,4,power_law_1.01,0.2893824100494385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,4,power_law_1.01,0.2914752006530762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,4,power_law_1.01,0.2855295896530151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,4,power_law_1.01,0.2967871904373169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,4,power_law_1.01,0.30495998859405515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,4,power_law_1.01,0.027270400524139406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,4,power_law_1.01,0.31505279541015624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,4,power_law_1.01,0.049491199851036075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,4,power_law_1.01,0.31140480041503904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,4,power_law_1.2,0.09070079922676086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,4,power_law_1.01,0.038867199420928956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,4,power_law_1.01,0.3353280067443848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,4,power_law_1.2,0.12363519668579101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,4,power_law_1.01,0.05238400101661682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,4,power_law_1.2,0.08120319843292237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.01,0.37105278968811034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,4,power_law_1.01,0.07124480009078979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,4,power_law_1.2,0.125600004196167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,4,power_law_1.2,0.11733759641647339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.01,0.4124288082122803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,4,power_law_1.01,0.08232960104942322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,4,power_law_1.2,0.19299839735031127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,4,power_law_1.2,0.14956799745559693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,4,power_law_1.2,0.10657279491424561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,4,power_law_1.01,0.10703999996185302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.01,0.47808637619018557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,4,power_law_1.2,0.17731839418411255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,4,power_law_1.2,0.17662080526351928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,4,power_law_1.01,0.11298559904098511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.01,0.5784704208374023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,4,power_law_1.2,0.23877758979797364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,4,power_law_1.2,0.20634241104125978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,4,power_law_1.01,0.1150015950202942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.01,0.636953592300415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,4,power_law_1.2,0.2481600046157837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,4,power_law_1.01,0.12097280025482178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.01,0.8708671569824219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,4,power_law_1.01,0.12170239686965942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,4,power_law_1.2,0.26481919288635253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.01,1.0641792297363282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,4,power_law_1.01,0.12085119485855103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,4,power_law_1.2,0.2748608112335205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.01,1.5613951683044434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,4,power_law_1.2,0.2823296070098877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,4,power_law_1.01,0.1261952042579651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.01,1.8041919708251952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,4,power_law_1.2,0.28728959560394285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,4,power_law_1.01,0.1339840054512024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,4,power_law_1.2,0.2807679891586304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,4,power_law_1.01,2.1765119552612306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,4,power_law_1.01,0.15150079727172852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,4,power_law_1.2,0.2980096101760864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,4,power_law_1.01,0.1428928017616272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,4,power_law_1.01,3.3446849822998046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,4,power_law_1.2,0.3219007968902588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,4,power_law_1.01,0.15084160566329957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,4,power_law_1.2,0.32739839553833006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,4,power_law_1.01,0.27809920310974123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,4,power_law_1.01,6.659705352783203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,4,power_law_1.2,0.33568639755249025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.01,0.3110975980758667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,4,power_law_1.2,0.36033918857574465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.01,0.24734079837799072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,4,power_law_1.2,0.4004799842834473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.01,0.2669440031051636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,4,power_law_1.2,0.44594559669494627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.01,0.3041343927383423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,4,power_law_1.2,0.5089983940124512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.01,0.3419264078140259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,4,power_law_1.2,0.6272192001342773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.01,0.4259903907775879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,4,power_law_1.2,0.7097152233123779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.01,0.5265600204467773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,4,power_law_1.2,0.9592127799987793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.01,0.6947968006134033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,4,power_law_1.2,1.1575488090515136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.01,0.8908543586730957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,4,power_law_1.2,1.6569215774536132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,4,power_law_1.01,1.1492608070373536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,4,power_law_1.2,2.0684736251831053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,4,power_law_1.01,1.7930816650390624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,4,power_law_1.2,2.5990591049194336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,4,power_law_1.01,3.2359424591064454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,4,power_law_1.2,4.320742416381836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,4,power_law_1.2,8.879052734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,4,power_law_1.2,0.11041920185089112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,4,power_law_1.2,0.13886719942092896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,4,power_law_1.2,0.09488000273704529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,4,power_law_1.2,0.11822719573974609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,4,power_law_1.2,0.2852159976959229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,4,power_law_1.2,0.1598912000656128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,4,power_law_1.2,0.3869184017181396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,4,power_law_1.2,0.19561599493026732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,4,power_law_1.2,0.4059328079223633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,4,power_law_1.2,0.2553152084350586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,4,power_law_1.2,0.42708477973937986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,4,power_law_1.2,0.25534720420837403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,4,power_law_1.2,0.26007680892944335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,4,power_law_1.2,0.4245632171630859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,4,power_law_1.2,0.28380799293518066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,4,power_law_1.2,0.4384255886077881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,4,power_law_1.2,0.027577599883079527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,4,power_law_1.2,0.28187520503997804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,2,balanced,0.08101333181063335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,4,power_law_1.2,0.463212776184082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,2,balanced,0.09404800335566203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,4,power_law_1.2,0.04764800071716309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,4,power_law_1.2,0.30675199031829836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,4,power_law_1.2,0.4768383979797363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,2,balanced,0.12104533116022746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,4,power_law_1.2,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,4,power_law_1.2,0.2983232021331787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,4,power_law_1.2,0.4742591857910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,2,balanced,0.1897760033607483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,4,power_law_1.2,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,2,balanced,0.30236266056696576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,4,power_law_1.2,0.30614399909973145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,4,power_law_1.2,0.49719038009643557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,2,balanced,0.5448373158772787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,4,power_law_1.2,0.06277120113372803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,4,power_law_1.2,0.32639360427856445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,4,power_law_1.2,0.5146880149841309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,2,balanced,0.5213386615117391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,4,power_law_1.2,0.07950080037117005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,4,power_law_1.2,0.3342720031738281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,2,balanced,0.5305440028508505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,4,power_law_1.2,0.5291967868804932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,2,balanced,0.5312480131785074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,4,power_law_1.2,0.10374399423599243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,4,power_law_1.2,0.327455997467041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,4,power_law_1.2,0.5767039775848388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,2,balanced,0.5446186860402426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,4,power_law_1.2,0.10629760026931763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,4,power_law_1.2,0.36912639141082765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,4,power_law_1.2,0.6111231803894043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,2,balanced,0.5170506636301676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,4,power_law_1.2,0.11455999612808228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.2,0.39908480644226074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,2,balanced,0.520309329032898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,4,power_law_1.2,0.6648384094238281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,4,power_law_1.2,0.11397759914398194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.2,0.42855038642883303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,2,balanced,0.5200053453445435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,4,power_law_1.2,0.7678400039672851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,4,power_law_1.2,0.12239999771118164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.2,0.5057216167449952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,2,balanced,0.5253066619237264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,4,power_law_1.2,0.9171135902404786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,4,power_law_1.2,0.1259328007698059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,2,balanced,0.5376586516698202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.2,0.5930816173553467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,4,power_law_1.2,1.0819904327392578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,4,power_law_1.2,0.12639360427856444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,2,balanced,0.5437920093536377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.2,0.6884223937988281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,4,power_law_1.2,0.12836480140686035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,4,power_law_1.2,1.3363903999328612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,2,balanced,0.5512053171793619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.2,0.9731136322021484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,4,power_law_1.2,0.14949120283126832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,4,power_law_1.2,1.5728639602661132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.2,1.0946687698364257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,2,balanced,0.574842651685079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,4,power_law_1.2,0.15488640069961548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.2,1.58155517578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,4,power_law_1.2,2.131315231323242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,2,balanced,0.5663573344548544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,4,power_law_1.2,0.15806080102920533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.2,1.9859455108642579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,4,power_law_1.2,2.695257568359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,4,power_law_1.2,0.2950527906417847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,2,balanced,0.6093279918034872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,4,power_law_1.2,2.5556671142578127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,4,power_law_1.2,3.1849599838256837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.2,0.31895039081573484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,4,power_law_1.2,3.573990249633789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,4,power_law_1.2,5.208211135864258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.2,0.2627968072891235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,2,balanced,0.6368639866511027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.2,0.2723520040512085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,4,power_law_1.2,7.260781097412109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,4,power_law_1.2,9.589798736572266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.2,0.3128319978713989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,2,balanced,0.7200906276702881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.2,0.35246078968048095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.2,0.4526847839355469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.2,0.5503935813903809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,2,balanced,0.7972213427225748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.2,0.7574016094207764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.2,1.0117568016052245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,4,power_law_1.2,1.180031967163086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,2,balanced,1.1360426743825276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,4,power_law_1.2,1.83636474609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,4,power_law_1.2,3.58458251953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,2,balanced,1.2568213144938152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,2,balanced,1.7699732780456543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,2,balanced,2.2281920115152993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,2,balanced,2.7107998530069985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,2,balanced,4.220111846923828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,2,balanced,0.07313066720962524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,2,balanced,0.06248533229033152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,2,balanced,0.09086400270462036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,2,balanced,0.07588799794514973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,2,balanced,0.1065013309319814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,2,balanced,0.09393067161242168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,2,balanced,0.12984533111254373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,2,balanced,0.16404799620310465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,2,balanced,0.20881599187850952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,2,balanced,0.2437600096066793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,2,balanced,0.3211839993794759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,2,balanced,0.3709119955698649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,2,balanced,0.31149866183598834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,2,balanced,0.3729120095570882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,2,balanced,0.31387199958165485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,2,balanced,0.3688960075378418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,2,balanced,0.31220799684524536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,2,balanced,0.3625280062357585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,2,balanced,0.30853867530822754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,2,balanced,0.35942399501800537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,2,balanced,0.3179733355840047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,2,balanced,0.3466613292694092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,2,balanced,0.31695467233657837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,2,balanced,0.3517226775487264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,2,balanced,0.325162669022878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,2,balanced,0.0236160010099411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,2,balanced,0.3571840127309163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,2,balanced,0.026709333062171936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,2,balanced,0.32285867134730023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,2,balanced,0.3556480010350545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,2,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,2,balanced,0.3280479907989502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,2,balanced,0.3620266517003377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,2,balanced,0.05693333347638448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,2,balanced,0.08630399902661641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,2,balanced,0.3343413273493449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,2,balanced,0.37698133786519367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,2,balanced,0.1368106702963511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,2,balanced,0.34489067395528156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,2,balanced,0.37878934542338055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,2,balanced,0.1395680010318756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,2,balanced,0.14271466930707297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,2,balanced,0.3708213170369466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,2,balanced,0.3911679983139038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,2,balanced,8.442709604899088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,2,balanced,0.14828266700108847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,2,balanced,0.1518186628818512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,2,balanced,0.39096001784006756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,2,balanced,0.41821332772572833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,2,balanced,0.15863999724388123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,2,balanced,0.15470932920773825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,2,balanced,0.4490400155385335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,2,balanced,0.43775999546051025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,2,balanced,0.16022933522860208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,2,balanced,0.1676479975382487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,2,balanced,0.4779946804046631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,2,balanced,0.46986134847005206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,2,balanced,0.17129600048065186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,2,balanced,0.17510932683944702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,2,balanced,0.18133866786956787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,2,balanced,0.5753653446833292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,2,balanced,0.5538773139317831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,2,balanced,0.23038933674494425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,2,balanced,0.2300800085067749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,2,balanced,0.6684373219807943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,2,balanced,0.6329439878463745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,2,balanced,0.3977866570154826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,2,balanced,0.4301333427429199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,2,balanced,0.9782026608784994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,2,balanced,0.968559980392456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,2,balanced,0.40753599007924396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,2,balanced,1.141594648361206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,2,balanced,0.43718934059143066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,2,balanced,1.1057066917419434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,2,balanced,0.49078933397928876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,2,balanced,1.6237173080444336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,2,balanced,0.5441493193308512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,2,balanced,1.5745654106140137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,2,balanced,2.0891040166219077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,2,balanced,0.8962773482004801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,2,balanced,2.0540000597635903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,2,balanced,1.015232006708781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,2,balanced,2.5558133125305176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,2,balanced,2.504042625427246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,2,balanced,1.3894559542338054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,2,balanced,3.9655946095784507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,2,balanced,1.9784746170043945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,2,balanced,3.937391916910807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,2,power_law_1.01,0.09567360281944275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,2,power_law_1.01,0.12433919906616211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,2,power_law_1.01,0.10355839729309083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,2,power_law_1.01,0.16331520080566406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,2,power_law_1.01,0.24296319484710693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,2,power_law_1.01,0.28091518878936766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,2,power_law_1.01,0.3874624013900757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,2,power_law_1.01,0.4096384048461914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,2,power_law_1.01,0.4206719875335693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,2,power_law_1.01,0.4403200149536133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,2,power_law_1.01,0.4450047969818115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,2,power_law_1.01,0.46117119789123534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,2,power_law_1.01,0.4696063995361328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,2,power_law_1.01,0.47235841751098634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,2,power_law_1.01,0.48775677680969237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,2,power_law_1.01,0.5011392116546631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,2,power_law_1.01,0.5017983913421631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,2,power_law_1.01,0.5470143795013428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,2,balanced,7.890421549479167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,2,power_law_1.01,0.580620813369751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,2,power_law_1.01,0.637824010848999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,2,power_law_1.01,0.7166336059570313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,2,power_law_1.01,0.8487423896789551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,2,power_law_1.01,1.005510425567627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,2,power_law_1.01,1.275551986694336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,2,balanced,3.8893601099650064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,2,power_law_1.01,1.534988784790039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,2,power_law_1.01,2.0725568771362304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,2,power_law_1.01,2.6758975982666016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,2,power_law_1.01,3.228799819946289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,2,power_law_1.01,4.770771026611328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,2,power_law_1.01,9.487993621826172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,2,balanced,7.815055847167969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,2,power_law_1.01,0.07411199808120728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,2,power_law_1.01,0.09488000273704529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,2,power_law_1.01,0.09359359741210938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,2,power_law_1.01,0.10497920513153076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,2,power_law_1.01,0.1553279995918274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,2,power_law_1.01,0.19379199743270875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,2,power_law_1.01,0.24297599792480468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,2,power_law_1.01,0.08799999952316284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,2,power_law_1.01,0.2543872117996216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,2,power_law_1.01,0.1114240050315857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,2,power_law_1.01,0.2621504068374634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,2,power_law_1.01,0.10721280574798583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,2,power_law_1.01,0.26922879219055174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,2,power_law_1.01,0.13555840253829957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,2,power_law_1.01,0.2750080108642578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,2,power_law_1.01,0.18702720403671264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,2,power_law_1.01,0.28501119613647463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,2,power_law_1.01,0.21857280731201173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,2,power_law_1.01,0.2880064010620117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,2,power_law_1.01,0.2799936056137085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,2,power_law_1.01,0.29920001029968263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,2,power_law_1.01,0.3026240110397339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,2,power_law_1.01,0.3084160089492798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,2,power_law_1.01,0.2989759922027588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,2,power_law_1.01,0.32152318954467773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,2,power_law_1.01,0.31135358810424807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,2,power_law_1.01,0.3239871978759766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,2,power_law_1.01,0.31536641120910647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,2,power_law_1.01,0.35790719985961916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,2,power_law_1.2,0.09722880125045777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,2,power_law_1.01,0.33435521125793455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,2,power_law_1.01,0.3985984086990356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,2,power_law_1.01,0.02240000069141388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,2,power_law_1.2,0.12042239904403687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,2,power_law_1.01,0.3235775947570801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,2,power_law_1.01,0.4557631969451904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,2,power_law_1.01,0.030073601007461547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,2,power_law_1.2,0.10119680166244507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,2,power_law_1.01,0.3303168058395386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,2,power_law_1.01,0.5291776180267334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,2,power_law_1.01,0.03086720108985901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,2,power_law_1.2,0.15932159423828124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,2,power_law_1.01,0.3426431894302368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,2,power_law_1.01,0.659449577331543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,2,power_law_1.01,0.04783360064029694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,2,power_law_1.2,0.20775039196014405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,2,power_law_1.01,0.35098240375518797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,2,power_law_1.01,0.7759168148040771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,2,power_law_1.01,0.06844800114631652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,2,power_law_1.2,0.2799616098403931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,2,power_law_1.01,0.35369598865509033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,2,power_law_1.01,1.036736011505127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,2,power_law_1.01,0.08026880025863647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,2,power_law_1.2,0.37348480224609376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,2,power_law_1.01,0.3835455894470215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,2,power_law_1.01,1.2765376091003418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,2,power_law_1.01,0.10469759702682495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,2,power_law_1.2,0.4038271903991699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.01,0.4080319881439209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,2,power_law_1.01,0.10837759971618652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,2,power_law_1.01,1.846847915649414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,2,power_law_1.2,0.4199999809265137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.01,0.4802239894866943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,2,power_law_1.01,0.11519360542297363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,2,power_law_1.01,2.2586496353149412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,2,power_law_1.2,0.42383360862731934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.01,0.5353151798248291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,2,power_law_1.01,0.11762559413909912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,2,power_law_1.2,0.4642496109008789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,2,power_law_1.01,2.8596607208251954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.01,0.6922815799713135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,2,power_law_1.01,0.1273344039916992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,2,power_law_1.2,0.45278081893920896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,2,power_law_1.01,4.360569763183594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.01,0.778272008895874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,2,power_law_1.01,0.1238976001739502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,2,power_law_1.2,0.4777791976928711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.01,1.0393407821655274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,2,power_law_1.01,0.12969599962234496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,2,power_law_1.2,0.48876161575317384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,2,power_law_1.01,8.840582275390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,2,power_law_1.01,0.1374783992767334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.01,1.3132672309875488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,2,power_law_1.2,0.49687681198120115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,2,power_law_1.01,0.1422144055366516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.01,1.8136127471923829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,2,power_law_1.2,0.5091775894165039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,2,power_law_1.01,0.14001280069351196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.01,2.3410688400268556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,2,power_law_1.2,0.5245567798614502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,2,power_law_1.01,0.1493183970451355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,2,power_law_1.2,0.5657087802886963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,2,power_law_1.01,2.9028736114501954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,2,power_law_1.01,0.16208640336990357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,2,power_law_1.2,0.6189568042755127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,2,power_law_1.01,4.358361434936524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.01,0.1750272035598755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,2,power_law_1.2,0.6818943977355957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.01,0.3398207902908325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,2,power_law_1.2,0.7570623874664306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,2,power_law_1.01,8.565580749511719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.01,0.3827264070510864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,2,power_law_1.2,0.9271743774414063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.01,0.31448960304260254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,2,power_law_1.2,1.0714240074157715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.01,0.3528127908706665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,2,power_law_1.2,1.3451647758483887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.01,0.4291520118713379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,2,power_law_1.2,1.6603967666625976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.01,0.5247807979583741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,2,power_law_1.2,2.315635108947754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.01,0.7181759834289551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,2,power_law_1.2,2.7840063095092775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.01,0.8769920349121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,2,power_law_1.2,3.2798526763916014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,2,power_law_1.01,1.0340415954589843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,2,power_law_1.2,5.118886566162109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,2,power_law_1.01,1.6110847473144532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,2,power_law_1.2,10.764678192138671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,2,power_law_1.01,3.2068801879882813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,2,power_law_1.2,0.07212160229682922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,2,power_law_1.2,0.09289600253105164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,2,power_law_1.2,0.07948799729347229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,2,power_law_1.2,0.11465599536895751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,2,power_law_1.2,0.08655359745025634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,2,power_law_1.2,0.14495999813079835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,2,power_law_1.2,0.113811194896698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,2,power_law_1.2,0.17529599666595458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,2,power_law_1.2,0.2351232051849365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,2,power_law_1.2,0.09811840057373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,2,power_law_1.2,0.24958720207214355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,2,power_law_1.2,0.13043839931488038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,2,power_law_1.2,0.2621056079864502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,2,power_law_1.2,0.166156804561615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,2,power_law_1.2,0.27143039703369143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,2,power_law_1.2,0.20062720775604248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,2,power_law_1.2,0.2764863967895508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,2,power_law_1.2,0.2781759977340698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,2,power_law_1.2,0.2979775905609131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,2,power_law_1.2,0.29577600955963135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,2,power_law_1.2,0.29339520931243895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,2,power_law_1.2,0.2950272083282471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,2,power_law_1.2,0.2990528106689453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,2,power_law_1.2,0.3036479949951172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,2,power_law_1.2,0.32321279048919677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,2,power_law_1.2,0.3140928030014038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,2,power_law_1.2,0.3228224039077759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,2,power_law_1.2,0.3247488021850586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,2,power_law_1.2,0.3507200002670288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,2,power_law_1.2,0.32229759693145754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,2,power_law_1.2,0.3706559896469116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,2,power_law_1.2,0.3345664024353027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,2,power_law_1.2,0.021913599967956544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,2,power_law_1.2,0.41315197944641113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,2,power_law_1.2,0.34420480728149416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,2,power_law_1.2,0.027616000175476073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,2,power_law_1.2,0.35477120876312257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,2,power_law_1.2,0.4768064022064209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,1,balanced,0.08966933687527974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,2,power_law_1.2,0.029164800047874452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,2,power_law_1.2,0.37348480224609376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,2,power_law_1.2,0.54202241897583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,1,balanced,0.10788800319035848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,2,power_law_1.2,0.04590719938278198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,2,power_law_1.2,0.40706558227539064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,1,balanced,0.1332319974899292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,2,power_law_1.2,0.6538943767547607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,1,balanced,0.20306134223937988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,2,power_law_1.2,0.06275839805603027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.2,0.4338240146636963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,2,power_law_1.2,0.7820543766021728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,1,balanced,0.3312586744626363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,2,power_law_1.2,0.07497599720954895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.2,0.5094016075134278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,2,power_law_1.2,1.0786751747131347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,1,balanced,0.5512640078862509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,1,balanced,0.5600853363672892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,2,power_law_1.2,0.10179840326309204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.2,0.5718207836151123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,2,power_law_1.2,1.3003199577331543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,1,balanced,0.5668960014979044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,2,power_law_1.2,0.10755840539932252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.2,0.7014527797698975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,2,power_law_1.2,1.9427776336669922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,1,balanced,0.5558559894561768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,2,power_law_1.2,0.11424000263214111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.2,0.8272319793701172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,1,balanced,0.5633013248443604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,2,power_law_1.2,2.3152128219604493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,2,power_law_1.2,0.11911040544509888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,1,balanced,0.5578933159510294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.2,1.0847935676574707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,2,power_law_1.2,2.834694480895996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,1,balanced,0.5554560025533041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,2,power_law_1.2,0.12416640520095826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.2,1.4322431564331055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,1,balanced,0.5680746634801229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,2,power_law_1.2,0.12609280347824098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,2,power_law_1.2,4.49288330078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.2,1.8003328323364258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,1,balanced,0.5703413486480713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,2,power_law_1.2,0.13074560165405275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.2,2.3262399673461913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,1,balanced,0.5708906650543213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,2,power_law_1.2,8.750393676757813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,2,power_law_1.2,0.13999359607696532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,2,power_law_1.2,2.9418752670288084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,2,power_law_1.2,0.13985919952392578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,2,power_law_1.2,0.14343680143356324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,2,power_law_1.2,4.686048126220703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,2,power_law_1.2,0.15176960229873657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,2,power_law_1.2,9.20931167602539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,2,power_law_1.2,0.16752640008926392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.2,0.17893120050430297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.2,0.3535615921020508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.2,0.4052735805511475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.2,0.33464961051940917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.2,0.37535359859466555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.2,0.4561791896820068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.2,0.5455167770385743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.2,0.7304319858551025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.2,0.9177663803100586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,2,power_law_1.2,1.1183296203613282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,2,power_law_1.2,1.6191680908203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,2,power_law_1.2,3.4428863525390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,1,balanced,0.07331199944019318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,1,balanced,0.0867146650950114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,1,balanced,0.11002666751543681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,1,balanced,0.15458133816719055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,1,balanced,0.241482675075531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,1,balanced,0.3906453450520833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,1,balanced,0.3843839963277181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,1,balanced,0.38688000043233234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,1,balanced,0.3858986695607503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,1,balanced,0.0993333359559377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,1,balanced,0.38300267855326336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,1,balanced,0.1109226644039154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,1,balanced,0.38550933202107746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,1,balanced,0.14417599638303122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,1,balanced,0.38975465297698975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,1,balanced,0.19592533508936563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,1,balanced,0.3953760067621867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,1,balanced,0.3012959957122803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,1,balanced,0.4015520016352336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,1,balanced,0.5026933352152506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,1,balanced,0.483573317527771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,1,balanced,0.4075253407160441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,1,balanced,0.473471999168396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,1,balanced,0.41702934106191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,1,balanced,0.4764853318532308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,1,balanced,0.4336053530375163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,1,balanced,0.479039986928304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,1,balanced,0.47223468621571857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,1,balanced,0.47618667284647626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,1,balanced,0.4802986780802409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,1,balanced,0.4845333496729533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,1,balanced,0.5069546699523926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,1,balanced,0.496234655380249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,1,balanced,0.5856746832529703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,1,balanced,0.5873706738154093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,1,balanced,0.5001066525777181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,1,balanced,0.023818666736284893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,1,balanced,0.5914453268051147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,1,balanced,0.02643200010061264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,1,balanced,0.523423989613851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,1,balanced,0.6374933322270712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,1,balanced,0.028736000259717304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,1,balanced,0.6530826489130656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,1,balanced,0.5217653512954712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,1,balanced,0.036864000062147774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,1,balanced,0.6632479826609293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,1,balanced,0.06206933160622915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,1,balanced,0.5603786706924438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,1,balanced,0.09670399626096089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,1,balanced,0.09800533453623454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,1,balanced,0.7303520043691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,1,balanced,0.8080106576283773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,1,balanced,0.5837706724802653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,1,balanced,0.10262933373451233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,1,balanced,0.10566932956377666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,1,balanced,0.1092693308989207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,1,balanced,0.7905546824137369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,1,balanced,0.653274655342102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,1,balanced,0.1142080028851827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,1,balanced,0.957914670308431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,1,balanced,0.12251200278600057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,1,balanced,0.7088960011800131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,1,balanced,0.9239412943522135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,1,balanced,0.1274186670780182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,1,balanced,0.14149333039919534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,1,balanced,0.13025599718093872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,1,balanced,0.8492266337076823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,1,balanced,0.13702932993570963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,1,balanced,1.0574026902516682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,1,balanced,1.4036426544189453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,1,balanced,0.1520746648311615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,1,balanced,0.15044800440470377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,1,balanced,0.9877599875132242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,1,balanced,0.16996800899505615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,1,balanced,1.5833652814229329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,1,balanced,0.19061867396036783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,1,balanced,1.7179360389709473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,1,balanced,1.5060532887776692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,1,balanced,0.2132800022761027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,1,balanced,0.33032000064849854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,1,balanced,1.8679626782735188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,1,balanced,1.7898292541503906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,1,balanced,0.38944534460703534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,1,balanced,2.4692907333374023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,1,balanced,0.38916265964508057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,1,balanced,2.6046719551086426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,1,balanced,2.5572640101114907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,1,balanced,0.4295733372370402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,1,balanced,3.1581761042277017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,1,balanced,3.3017279307047525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,1,balanced,3.350794792175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,1,balanced,0.6935253143310547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,1,balanced,0.8050080140431722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,1,balanced,3.883082707722982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,1,balanced,4.044602711995442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,1,balanced,4.1398881276448565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,1,balanced,1.0721173286437988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,1,balanced,6.472821553548177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,1,balanced,6.03437868754069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,1,balanced,6.247018814086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,1,balanced,1.5579946835835774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,1,balanced,12.761039733886719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,1,balanced,11.844751993815104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,1,balanced,12.19485855102539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,1,balanced,3.044661204020182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,1,power_law_1.01,0.08389120101928711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,1,power_law_1.01,0.0998528003692627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,1,power_law_1.01,0.1050112009048462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,1,power_law_1.01,0.1756991982460022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,1,power_law_1.01,0.23138558864593506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,1,power_law_1.01,0.29665920734405515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,1,power_law_1.01,0.40543360710144044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,1,power_law_1.01,0.06920959949493408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,1,power_law_1.01,0.42414078712463377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,1,power_law_1.01,0.08136320114135742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,1,power_law_1.01,0.09489920139312744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,1,power_law_1.01,0.44077439308166505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,1,power_law_1.01,0.09068160057067871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,1,power_law_1.01,0.10531840324401856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,1,power_law_1.01,0.02136960029602051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,1,power_law_1.01,0.4483776092529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,1,power_law_1.01,0.13305599689483644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,1,power_law_1.01,0.11664639711380005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,1,power_law_1.01,0.02455040067434311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,1,power_law_1.01,0.4773568153381348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,1,power_law_1.01,0.1764415979385376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,1,power_law_1.01,0.16881279945373534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,1,power_law_1.01,0.026943999528884887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,1,power_law_1.01,0.4852287769317627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,1,power_law_1.01,0.22272000312805176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,1,power_law_1.01,0.22551040649414061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,1,power_law_1.01,0.03646079897880554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,1,power_law_1.01,0.4923136234283447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,1,power_law_1.01,0.2855295896530151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,1,power_law_1.01,0.27384960651397705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,1,power_law_1.01,0.05786240100860596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,1,power_law_1.01,0.4983168125152588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,1,power_law_1.01,0.2949120044708252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,1,power_law_1.01,0.3504640102386475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,1,power_law_1.01,0.08511360287666321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,1,power_law_1.01,0.5171775817871094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,1,power_law_1.01,0.30191359519958494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,1,power_law_1.01,0.37160320281982423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,1,power_law_1.01,0.10005120038986207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,1,power_law_1.01,0.5229631900787354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,1,power_law_1.01,0.3161535978317261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,1,power_law_1.01,0.3894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,1,power_law_1.01,0.5491327762603759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,1,power_law_1.01,0.10984959602355956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,1,power_law_1.01,0.33000960350036623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,1,power_law_1.01,0.3897792100906372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,1,power_law_1.01,0.6125696182250977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,1,power_law_1.01,0.11477119922637939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,1,power_law_1.01,0.4070591926574707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,1,power_law_1.01,0.34337921142578126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,1,power_law_1.01,0.6437695980072021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,1,power_law_1.01,0.1197119951248169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,1,power_law_1.01,0.415667200088501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,1,power_law_1.01,0.337990403175354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,1,power_law_1.01,0.1273151993751526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,1,power_law_1.01,0.42476801872253417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,1,power_law_1.01,0.3516671895980835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,1,power_law_1.01,0.13528319597244262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,1,power_law_1.01,0.4371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,1,power_law_1.01,0.3724544048309326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,1,power_law_1.01,0.14252159595489503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,1,power_law_1.01,0.4500288009643555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,1,power_law_1.01,0.38178560733795164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,1,power_law_1.01,0.153766405582428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,1,power_law_1.01,0.4035647869110107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,1,power_law_1.01,0.4585087776184082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,1,power_law_1.01,0.1478592038154602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,1,power_law_1.01,0.45698561668396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,1,power_law_1.01,0.4763967990875244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,1,power_law_1.01,0.15543680191040038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,1,power_law_1.01,0.4933119773864746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,1,power_law_1.01,0.5288576126098633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,1,power_law_1.01,0.17065600156784058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.01,0.580620813369751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,1,power_law_1.01,0.59269118309021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,1,power_law_1.01,0.17144960165023804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.01,0.6778560161590577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.01,0.19177600145339965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,1,power_law_1.01,0.6815936088562011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.01,0.7747519969940185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.01,0.21994240283966066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,1,power_law_1.01,0.8620672225952148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.01,0.9686719894409179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.01,0.2740288019180298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,1,power_law_1.01,1.049120044708252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.01,0.4413887977600098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.01,1.1568384170532227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,1,power_law_1.01,1.403001594543457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.01,0.5207551956176758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,1,power_law_1.01,1.780998420715332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.01,1.5514431953430177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.01,0.47087998390197755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.01,1.9439231872558593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,1,power_law_1.01,2.4943424224853517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.01,0.5622719764709473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.01,2.7205631256103517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,1,power_law_1.01,3.2183040618896483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.01,0.7377088069915771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.01,3.4943359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,1,power_law_1.01,3.9519615173339844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.01,0.9236096382141114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.01,4.269184112548828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,1,power_law_1.01,6.131379318237305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.01,1.1004287719726562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,1,power_law_1.01,6.578765106201172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,1,power_law_1.01,11.98946533203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,1,power_law_1.01,1.661408042907715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,1,power_law_1.01,12.781407928466797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,1,power_law_1.01,3.1654272079467773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,1,power_law_1.01,0.7474239826202392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,1,power_law_1.01,0.8315263748168945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,1,power_law_1.01,1.0124159812927247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,1,power_law_1.01,1.2091903686523438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,1,power_law_1.2,0.08191999793052673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,1,power_law_1.01,1.568000030517578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,1,power_law_1.2,0.10339200496673584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,1,power_law_1.01,1.9609535217285157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,1,power_law_1.2,0.10683519840240478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,1,power_law_1.2,0.0688704013824463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,1,power_law_1.2,0.09264000058174134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,1,power_law_1.01,2.6607423782348634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,1,power_law_1.2,0.17099519968032836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,1,power_law_1.2,0.08328319787979126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,1,power_law_1.2,0.10426880121231079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,1,power_law_1.2,0.2089087963104248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,1,power_law_1.01,3.4148670196533204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,1,power_law_1.2,0.09111040234565734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,1,power_law_1.2,0.11087360382080078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,1,power_law_1.2,0.2631808042526245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,1,power_law_1.01,4.149273681640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,1,power_law_1.2,0.13041919469833374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,1,power_law_1.2,0.1664512038230896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,1,power_law_1.2,0.3948800086975098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,1,power_law_1.01,6.385811233520508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,1,power_law_1.2,0.16643199920654297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,1,power_law_1.2,0.41137280464172366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,1,power_law_1.2,0.20707199573516846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,1,power_law_1.2,0.20620160102844237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,1,power_law_1.2,0.24982399940490724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,1,power_law_1.2,0.415231990814209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,1,power_law_1.01,12.363136291503906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,1,power_law_1.2,0.2836416006088257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,1,power_law_1.2,0.4520063877105713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,1,power_law_1.2,0.3486464023590088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,1,power_law_1.2,0.2975872039794922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,1,power_law_1.2,0.4478464126586914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,1,power_law_1.2,0.368831992149353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,1,power_law_1.2,0.30874879360198976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,1,power_law_1.2,0.4763455867767334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,1,power_law_1.2,0.37230079174041747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,1,power_law_1.2,0.3161920070648193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,1,power_law_1.2,0.48463997840881345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,1,power_law_1.2,0.39009919166564944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,1,power_law_1.2,0.32677760124206545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,1,power_law_1.2,0.5009984016418457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,1,power_law_1.2,0.4074751853942871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,1,power_law_1.2,0.3469504117965698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,1,power_law_1.2,0.5265984058380127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,1,power_law_1.2,0.4131328105926514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,1,power_law_1.2,0.36191999912261963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,1,power_law_1.2,0.5329728126525879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,1,power_law_1.2,0.4210944175720215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,1,power_law_1.2,0.3675328016281128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,1,power_law_1.2,0.5671552181243896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,1,power_law_1.2,0.4413760185241699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,1,power_law_1.2,0.3835200071334839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,1,power_law_1.2,0.6169087886810303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,1,power_law_1.2,0.4582655906677246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,1,power_law_1.2,0.3841344118118286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,1,power_law_1.2,0.6758336067199707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,1,power_law_1.2,0.46645121574401854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,1,power_law_1.2,0.4244351863861084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,1,power_law_1.2,0.7648640155792237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,1,power_law_1.2,0.49674239158630373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,1,power_law_1.2,0.4745664119720459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,1,power_law_1.2,0.8446080207824707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,1,power_law_1.2,0.5607999801635742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,1,power_law_1.2,0.519865608215332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,1,power_law_1.2,1.0390975952148438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.2,0.5928768157958985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,1,power_law_1.2,0.6175871849060058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,1,power_law_1.2,1.2167231559753418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.2,0.7051072120666504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,1,power_law_1.2,0.7026879787445068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,1,power_law_1.2,1.581753635406494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.2,0.7944191932678223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,1,power_law_1.2,0.8919551849365235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,1,power_law_1.2,1.9345983505249023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.2,0.992147159576416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,1,power_law_1.2,1.0699775695800782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.2,1.1919487953186034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,1,power_law_1.2,2.6715839385986326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,1,power_law_1.2,1.4372991561889648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.2,1.5764736175537108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,1,power_law_1.2,3.3984447479248048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,1,power_law_1.2,1.7900543212890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.2,1.9650047302246094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,1,power_law_1.2,4.134656143188477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,1,power_law_1.2,2.5250816345214844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.2,2.7330432891845704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,1,power_law_1.2,6.314284896850586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,1,power_law_1.2,3.2362945556640623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.2,3.499520111083984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,1,power_law_1.2,12.194489288330079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,1,power_law_1.2,3.9739585876464845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.2,4.265663909912109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,1,power_law_1.2,6.144620895385742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,1,power_law_1.2,6.5773567199707035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,1,power_law_1.2,11.971942138671874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,1,power_law_1.2,12.776646423339844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,1,power_law_1.2,0.021689599752426146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,1,power_law_1.2,0.023296000063419343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,1,power_law_1.2,0.026694399118423463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,1,power_law_1.2,0.03640959858894348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,1,power_law_1.2,0.055270397663116456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,1,power_law_1.2,0.07965440154075623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,1,power_law_1.2,0.10269440412521362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,1,power_law_1.2,0.11051520109176635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,1,power_law_1.2,0.11248639822006226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,1,power_law_1.2,0.11425919532775879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,1,power_law_1.2,0.1282047986984253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,1,power_law_1.2,0.13560320138931276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,1,power_law_1.2,0.13831679821014403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,1,power_law_1.2,0.15363199710845948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,1,power_law_1.2,0.14662400484085084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,1,power_law_1.2,0.15735679864883423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,1,power_law_1.2,0.17195520401000977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,1,power_law_1.2,0.1786239981651306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.2,0.19500160217285156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.2,0.22602241039276122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.2,0.2824959993362427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.2,0.45610880851745605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.2,0.5301055908203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.2,0.48524160385131837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.2,0.5706495761871337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.2,0.7566143989562988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.2,0.9374527931213379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,8,balanced,0.03388266762097677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,8,balanced,0.03850133220354716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.2,1.1358016014099122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,8,balanced,0.042992000778516136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,8,balanced,0.03390933324893316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,1,power_law_1.2,1.6787776947021484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,8,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,8,balanced,0.03387733300526937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,8,balanced,0.03839466720819473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,8,balanced,0.06403733293215434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,8,balanced,0.035631999373435974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,8,balanced,0.04081066697835922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,1,power_law_1.2,3.179327964782715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,8,balanced,0.06439466774463654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,8,balanced,0.03902933249870936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,8,balanced,0.04978133241335551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,8,balanced,0.04541866481304169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,8,balanced,0.067071999112765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,8,balanced,0.057392001152038574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,8,balanced,0.05120533208052317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,8,balanced,0.06950933237870534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,8,balanced,0.057376002271970115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,8,balanced,0.052629331747690834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,8,balanced,0.06698666512966156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,8,balanced,0.05750933289527893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,8,balanced,0.049925332268079124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,8,balanced,0.07166400055090587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,8,balanced,0.05529066423575083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,8,balanced,0.05504000186920166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,8,balanced,0.06549866497516632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,8,balanced,0.0514933317899704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,8,balanced,0.05502399802207947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,8,balanced,0.07064533233642578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,8,balanced,0.05115733544031779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,8,balanced,0.05470933516820272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,8,balanced,0.0672106643517812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,8,balanced,0.052815998593966164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,8,balanced,0.05611200133959452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,8,balanced,0.053183997670809426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,8,balanced,0.057034666339556374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,8,balanced,0.06713066498438518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,8,balanced,0.05464000006516775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,8,balanced,0.057162667314211525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,8,balanced,0.07319466769695282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,8,balanced,0.05316799879074097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,8,balanced,0.05913599828879038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,8,balanced,0.0710453341404597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,8,balanced,0.05765333275000254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,8,balanced,0.06295466423034668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,8,balanced,0.07099199791749318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,8,balanced,0.05868266522884369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,8,balanced,0.06074133515357971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,8,balanced,0.07398400207360585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,8,balanced,0.05857066810131073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,8,balanced,0.06081599990526835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,8,balanced,0.0624533345301946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,8,balanced,0.06578133503595988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,8,balanced,0.07406400144100189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,8,balanced,0.06508799890677135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,8,balanced,0.06897066533565521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,8,balanced,0.08077333370844524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,8,balanced,0.07794133325417836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,8,balanced,0.07188799977302551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,8,balanced,0.08557333548863728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,8,balanced,0.08098666866620381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,8,balanced,0.08172266681989034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,8,balanced,0.10290666421254475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,8,balanced,0.10799466570218404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,8,balanced,0.10355732838312785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,8,balanced,0.1220266620318095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,8,balanced,0.11980799833933513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,8,balanced,0.11180266737937927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,8,balanced,0.16100800037384033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,8,balanced,0.15441600481669107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,8,balanced,0.14038933316866556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,8,balanced,0.19814932346343994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,8,balanced,0.18825600544611612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,8,balanced,0.16845333576202393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,8,balanced,0.27739200989405316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,8,balanced,0.2606719930966695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,8,balanced,0.22772266467412314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,8,balanced,0.3616960048675537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,8,balanced,0.32524800300598145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,8,balanced,0.29500800371170044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,8,balanced,0.44657599925994873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,8,balanced,0.39667201042175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,8,balanced,0.356661319732666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,8,balanced,0.019248000035683315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,8,balanced,0.02160000056028366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,8,balanced,0.022229333718617756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,8,balanced,0.6859040260314941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,8,balanced,0.5964320103327433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,8,balanced,0.022069332500298817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,8,balanced,0.024282666544119518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,8,balanced,0.0271573339899381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,8,balanced,0.030447999636332195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,8,balanced,0.030234667162100475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,8,balanced,0.03808533400297165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,8,balanced,0.03823466598987579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,8,balanced,0.034389334420363106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,8,balanced,0.046426668763160706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,8,balanced,0.04623466730117798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,8,balanced,0.04576533536116282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,8,balanced,0.05706666906674703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,8,balanced,0.05764266848564148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,8,balanced,0.05295466880003611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,8,balanced,0.056015998125076294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,8,balanced,0.0650186687707901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,8,balanced,0.06757333377997081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,8,balanced,0.535151998202006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,8,balanced,0.07668266693751018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,8,balanced,0.08575466275215149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,8,balanced,0.10341333349545796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,8,balanced,0.1492800017197927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,8,balanced,0.17287999391555786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,8,balanced,1.306874672571818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,8,balanced,1.1270346641540527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,8,balanced,0.24623467524846396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,8,balanced,0.31886933247248334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,8,balanced,0.3903573354085286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,8,balanced,1.0065386295318604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,8,balanced,0.6117920080820719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,8,balanced,1.1974559624989827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,8,power_law_1.01,0.04368639886379242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,8,power_law_1.01,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,8,power_law_1.01,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,8,power_law_1.01,0.04431360065937042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,8,power_law_1.01,0.042745599150657655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,8,power_law_1.01,0.04456959962844849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,8,power_law_1.01,0.04402559995651245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,8,power_law_1.01,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,8,power_law_1.01,0.05134720206260681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,8,power_law_1.01,0.04618239998817444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,8,power_law_1.01,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,8,power_law_1.01,0.05082240104675293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,8,power_law_1.01,0.050995200872421265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,8,power_law_1.01,0.06436480283737182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,8,power_law_1.01,0.051820802688598636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,8,power_law_1.01,0.06656000018119812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,8,power_law_1.01,0.052883201837539674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,8,power_law_1.01,0.06588799953460693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,8,power_law_1.01,0.05326719880104065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,8,power_law_1.01,0.06738560199737549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,8,power_law_1.01,0.05436800122261047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,8,power_law_1.01,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,8,power_law_1.01,0.053523200750350955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,8,power_law_1.01,0.06694399714469909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,8,power_law_1.01,0.05539199709892273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,8,power_law_1.01,0.06785920262336731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,8,power_law_1.01,0.05865600109100342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,8,power_law_1.01,0.06779519915580749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,8,power_law_1.01,0.06316159963607788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,8,power_law_1.01,0.06816639900207519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,8,power_law_1.01,0.07137920260429383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,8,power_law_1.01,0.0652671992778778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,8,power_law_1.01,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,8,power_law_1.01,0.07040640115737914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,8,power_law_1.01,0.07288320064544677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,8,power_law_1.01,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,8,power_law_1.01,0.08060160279273987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,8,power_law_1.01,0.08900480270385742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,8,power_law_1.01,0.08606719970703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,8,power_law_1.01,0.10394879579544067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,8,power_law_1.01,0.09557759761810303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,8,power_law_1.01,0.13258880376815796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,8,power_law_1.01,0.10592639446258545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,8,power_law_1.01,0.1575551986694336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,8,power_law_1.01,0.13541760444641113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,8,power_law_1.01,0.2099839925765991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,8,power_law_1.01,0.15337599515914918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,8,power_law_1.01,0.27730560302734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,8,power_law_1.01,0.20351359844207764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,8,power_law_1.01,0.42987518310546874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,8,power_law_1.01,0.26888959407806395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,8,power_law_1.01,0.5454016208648682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,8,power_law_1.01,0.04723199903964996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,8,power_law_1.01,0.32635519504547117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,8,power_law_1.01,0.6116223812103272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,8,power_law_1.01,0.04121600091457367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,8,power_law_1.01,0.44917759895324705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,8,power_law_1.01,0.9991231918334961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,8,power_law_1.01,0.04314239919185638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,8,power_law_1.01,0.5286143779754638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,8,power_law_1.01,0.04705280065536499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,8,power_law_1.01,1.7530624389648437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,8,power_law_1.01,0.860870361328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,8,power_law_1.01,0.047295999526977536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,8,power_law_1.01,0.05424000024795532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,8,power_law_1.01,1.5160256385803224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,8,power_law_1.01,0.054739201068878175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,8,power_law_1.01,0.05466880202293396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,8,power_law_1.01,0.054662400484085084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,8,power_law_1.01,0.05594239830970764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,8,power_law_1.01,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,8,power_law_1.01,0.05665919780731201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,8,power_law_1.01,0.05702400207519531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,8,power_law_1.01,0.05860480070114136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,8,power_law_1.01,0.06058239936828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,8,power_law_1.01,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,8,power_law_1.01,0.06656640172004699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,8,power_law_1.01,0.07102720141410827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,8,power_law_1.01,0.07806079983711242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,8,power_law_1.01,0.09344639778137206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,8,power_law_1.01,0.10465919971466064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,8,power_law_1.01,0.1265663981437683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,8,power_law_1.01,0.1551743984222412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,8,power_law_1.01,0.2062527894973755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,8,power_law_1.01,0.2721343994140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,8,power_law_1.01,0.3754048109054565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,8,power_law_1.01,0.4721536159515381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,8,power_law_1.01,0.5746047973632813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,8,power_law_1.01,0.8229248046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,8,power_law_1.01,1.7589248657226562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,8,power_law_1.01,0.017958399653434754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,8,power_law_1.01,0.01934719979763031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,8,power_law_1.01,0.019225600361824035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,8,power_law_1.01,0.019993600249290467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,8,power_law_1.01,0.021766400337219237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,8,power_law_1.01,0.02366719990968704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,8,power_law_1.01,0.026124799251556398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,8,power_law_1.01,0.024889600276947022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,8,power_law_1.01,0.031564798951148984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,8,power_law_1.01,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,8,power_law_1.01,0.031071999669075014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,8,power_law_1.01,0.04454399943351746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,8,power_law_1.01,0.04526079893112182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,8,power_law_1.01,0.04340479969978332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,8,power_law_1.01,0.050483202934265135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,8,power_law_1.01,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,8,power_law_1.01,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,8,power_law_1.01,0.05201280117034912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,8,power_law_1.01,0.048825600743293764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,8,power_law_1.01,0.05578879714012146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,8,power_law_1.01,0.061273598670959474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,8,power_law_1.01,0.0726144015789032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,8,power_law_1.01,0.0874176025390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,8,power_law_1.01,0.11299200057983398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,8,power_law_1.01,0.13642239570617676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,8,power_law_1.01,0.18769919872283936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,8,power_law_1.01,0.24332160949707032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,8,power_law_1.01,0.292902398109436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,8,power_law_1.01,0.4470975875854492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,8,power_law_1.01,0.9431424140930176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,8,power_law_1.2,0.044064000248909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,8,power_law_1.2,0.03904640078544617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,8,power_law_1.2,0.040505599975585935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,8,power_law_1.2,0.04374400079250336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,8,power_law_1.2,0.04790399968624115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,8,power_law_1.2,0.04403199851512909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,8,power_law_1.2,0.06056960225105286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,8,power_law_1.2,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,8,power_law_1.2,0.0640064001083374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,8,power_law_1.2,0.034227201342582704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,8,power_law_1.2,0.06784639954566955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,8,power_law_1.2,0.0658623993396759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,8,power_law_1.2,0.040268799662590025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,8,power_law_1.2,0.0705344021320343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,8,power_law_1.2,0.04101119935512543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,8,power_law_1.2,0.0667519986629486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,8,power_law_1.2,0.04832639992237091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,8,power_law_1.2,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,8,power_law_1.2,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,8,power_law_1.2,0.06688640117645264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,8,power_law_1.2,0.052198398113250735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,8,power_law_1.2,0.07034239768981934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,8,power_law_1.2,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,8,power_law_1.2,0.07340160012245178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,8,power_law_1.2,0.05160319805145264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,8,power_law_1.2,0.07441920042037964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,8,power_law_1.2,0.052025598287582395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,8,power_law_1.2,0.07763839960098266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,8,power_law_1.2,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,8,power_law_1.2,0.08221439719200134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,8,power_law_1.2,0.052876800298690796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,8,power_law_1.2,0.09007999897003174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,8,power_law_1.2,0.05655679702758789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,8,power_law_1.2,0.10271999835968018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,8,power_law_1.2,0.059935998916625974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,8,power_law_1.2,0.11590399742126464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,8,power_law_1.2,0.0615231990814209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,8,power_law_1.2,0.13178240060806273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,8,power_law_1.2,0.06408960223197938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,8,power_law_1.2,0.07304319739341736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,8,power_law_1.2,0.1696128010749817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,8,power_law_1.2,0.08258559703826904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,8,power_law_1.2,0.22071681022644044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,8,power_law_1.2,0.0969983994960785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,8,power_law_1.2,0.2930304050445557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,8,power_law_1.2,0.0457152009010315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,8,power_law_1.2,0.10595200061798096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,8,power_law_1.2,0.4273215770721436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,8,power_law_1.2,0.03898879885673523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,8,power_law_1.2,0.14237439632415771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,8,power_law_1.2,0.49584641456604006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,8,power_law_1.2,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,8,power_law_1.2,0.1700543999671936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,8,power_law_1.2,0.04319359958171844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,8,power_law_1.2,0.6875840187072754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,8,power_law_1.2,0.04707199931144714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,8,power_law_1.2,0.22697598934173585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,8,power_law_1.2,1.1201472282409668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,8,power_law_1.2,0.053491199016571046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,8,power_law_1.2,0.3073535919189453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,8,power_law_1.2,2.0349504470825197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,8,power_law_1.2,0.05449600219726562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,8,power_law_1.2,0.47002239227294923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,8,power_law_1.2,0.05484799742698669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,8,power_law_1.2,0.5920896053314209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,8,power_law_1.2,0.055936002731323244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,8,power_law_1.2,0.7566847801208496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,8,power_law_1.2,0.053887999057769774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,8,power_law_1.2,1.4187264442443848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,8,power_law_1.2,0.05795199871063232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,8,power_law_1.2,0.056883198022842404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,8,power_law_1.2,2.934259223937988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,8,power_law_1.2,0.05735039710998535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,8,power_law_1.2,0.05782399773597717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,8,power_law_1.2,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,8,power_law_1.2,0.06335999965667724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,8,power_law_1.2,0.06649600267410279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,8,power_law_1.2,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,8,power_law_1.2,0.07817599773406983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,8,power_law_1.2,0.09468160271644592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,8,power_law_1.2,0.11258879899978638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,8,power_law_1.2,0.1569216012954712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,8,power_law_1.2,0.17712639570236205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,8,power_law_1.2,0.25911679267883303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,8,power_law_1.2,0.2755903959274292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,8,power_law_1.2,0.4708415985107422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,8,power_law_1.2,0.6284736156463623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,8,power_law_1.2,0.7010176181793213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,8,power_law_1.2,1.065817642211914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,8,power_law_1.2,2.8316736221313477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,8,power_law_1.2,0.018585599958896637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,8,power_law_1.2,0.01935359984636307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,8,power_law_1.2,0.01961600035429001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,8,power_law_1.2,0.02022400051355362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,8,power_law_1.2,0.021216000616550445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,8,power_law_1.2,0.022841599583625794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,8,power_law_1.2,0.0262719988822937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,8,power_law_1.2,0.025536000728607178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,8,power_law_1.2,0.030995199084281923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,8,power_law_1.2,0.0328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,8,power_law_1.2,0.02968960106372833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,8,power_law_1.2,0.044710400700569156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,8,power_law_1.2,0.04488320052623749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,8,power_law_1.2,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,8,power_law_1.2,0.052665597200393675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,8,power_law_1.2,0.05138559937477112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,8,power_law_1.2,0.04832639992237091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,8,power_law_1.2,0.05002239942550659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,8,power_law_1.2,0.05112959742546082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,8,power_law_1.2,0.057740801572799684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,8,power_law_1.2,0.061561602354049685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,8,power_law_1.2,0.07411839962005615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,8,power_law_1.2,0.08885120153427124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,8,power_law_1.2,0.11964160203933716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,8,power_law_1.2,0.1493183970451355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,8,power_law_1.2,0.20798079967498778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,8,power_law_1.2,0.2608576059341431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,8,power_law_1.2,0.3310976028442383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,8,power_law_1.2,0.4878335952758789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,8,power_law_1.2,1.0448448181152343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,4,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,4,balanced,0.03634133438269297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,4,balanced,0.039317332208156586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,4,balanced,0.03437866767247518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,4,balanced,0.049082666635513306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,4,balanced,0.03490666548411051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,4,balanced,0.06750399867693584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,4,balanced,0.03745600084463755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,4,balanced,0.07170133292675018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,4,balanced,0.044821331898371376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,4,balanced,0.06806933383146922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,4,balanced,0.053616002202034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,4,balanced,0.06788800160090129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,4,balanced,0.055120001236597695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,4,balanced,0.0689279983441035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,4,balanced,0.055349335074424744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,4,balanced,0.06876799960931142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,4,balanced,0.056234667698542275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,4,balanced,0.06715733309586842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,4,balanced,0.054901331663131714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,4,balanced,0.07028266787528992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,4,balanced,0.054746667544047035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,4,balanced,0.06946133573849995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,4,balanced,0.05495999753475189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,4,balanced,0.0691786656777064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,4,balanced,0.05677866439024607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,4,balanced,0.07576533158620198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,4,balanced,0.0592853327592214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,4,balanced,0.07308266560236613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,4,balanced,0.05724266668160757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,4,balanced,0.06308799982070923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,4,balanced,0.07550399998823802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,4,balanced,0.06458133459091187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,4,balanced,0.07950933277606964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,4,balanced,0.06509866813818614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,4,balanced,0.08372267087300618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,4,balanced,0.06976533432801564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,4,balanced,0.03692800054947535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,4,balanced,0.041034666200478874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,4,balanced,0.0737013320128123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,4,balanced,0.04508799811204275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,4,balanced,0.09027733405431111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,4,balanced,0.053823997577031456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,4,balanced,0.06711466610431671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,4,balanced,0.09292266766230266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,4,balanced,0.06365866462389629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,4,balanced,0.09763733545939128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,4,balanced,0.06276800235112508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,4,balanced,0.06334400177001953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,4,balanced,0.10339732964833577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,4,balanced,0.06392000118891399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,4,balanced,0.06345599889755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,4,balanced,0.12147733569145203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,4,balanced,0.06333333253860474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,4,balanced,0.13447466492652893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,4,balanced,0.0645066648721695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,4,balanced,0.0651093324025472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,4,balanced,0.06464533507823944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,4,balanced,0.06926399966080983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,4,balanced,0.1529813309510549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,4,balanced,0.13371200362841287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,4,balanced,0.07189333438873291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,4,balanced,0.07297599812348683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,4,balanced,0.07905599971612294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,4,balanced,0.2101973295211792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,4,balanced,0.08236800134181976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,4,balanced,0.17561600605646768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,4,balanced,0.09596266349156697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,4,balanced,0.1055519978205363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,4,balanced,0.26572267214457196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,4,balanced,0.14654933412869772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,4,balanced,0.2172373334566752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,4,balanced,0.16513066490491232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,4,balanced,0.3850613435109456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,4,balanced,0.23342400789260864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,4,balanced,0.30803199609120685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,4,balanced,0.2946079969406128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,4,balanced,0.5077493190765381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,4,balanced,0.41919998327891034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,4,balanced,0.39485331376393634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,4,balanced,0.017770666629076004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,4,balanced,0.0198186660806338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,4,balanced,0.02176533391078313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,4,balanced,0.6299413442611694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,4,balanced,0.022277332842350006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,4,balanced,0.024400000770886738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,4,balanced,0.02608533451954524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,4,balanced,0.027829334139823914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,4,balanced,0.02844800055027008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,4,balanced,0.5418560107549032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,4,balanced,0.030207999050617218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,4,balanced,0.03013866643110911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,4,balanced,0.030117332935333252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,4,balanced,0.03839466720819473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,4,balanced,0.03728000074625015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,4,balanced,0.03661333272854487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,4,balanced,0.048512001832326256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,4,balanced,0.04970133304595947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,4,balanced,0.05014933149019877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,4,balanced,0.061066667238871254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,4,balanced,0.06391466657320659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,4,balanced,0.48233067989349365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,4,balanced,0.6612480084101359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,4,balanced,0.069541335105896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,4,balanced,0.07708266874154408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,4,balanced,0.9727679888407389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,4,balanced,0.08341866731643677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,4,balanced,0.09055466453234355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,4,balanced,0.13853333393732706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,4,balanced,0.1530453364054362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,4,balanced,1.0143146514892578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,4,balanced,0.7305119832356771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,4,balanced,0.21628799041112265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,4,balanced,0.27771200736363727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,4,balanced,1.871946652730306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,4,balanced,0.34007465839385986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,4,balanced,1.940656026204427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,4,balanced,0.525322675704956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,4,balanced,1.383519967397054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,4,balanced,1.0221333503723145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,4,power_law_1.01,0.03768959939479828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,4,power_law_1.01,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,4,power_law_1.01,0.0378495991230011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,4,power_law_1.01,0.04656639993190766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,4,power_law_1.01,0.05073919892311096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,4,power_law_1.01,0.06438400149345398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,4,power_law_1.01,0.06551039814949036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,4,power_law_1.01,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,4,power_law_1.01,0.06709759831428527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,4,power_law_1.01,0.06467199921607972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,4,power_law_1.01,0.03840000033378601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,4,power_law_1.01,0.06912639737129211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,4,power_law_1.01,0.03692159950733185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,4,power_law_1.01,0.06865919828414917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,4,power_law_1.01,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,4,power_law_1.01,0.06944000124931335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,4,power_law_1.01,0.04000000059604645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,4,power_law_1.01,0.06821119785308838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,4,power_law_1.01,0.04359039962291718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,4,power_law_1.01,0.05160319805145264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,4,power_law_1.01,0.07455360293388366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,4,power_law_1.01,0.04967679977416992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,4,power_law_1.01,0.07397119998931885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,4,power_law_1.01,0.053260797262191774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,4,power_law_1.01,0.07612159848213196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,4,power_law_1.01,0.05261440277099609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,4,power_law_1.01,0.08197759985923767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,4,power_law_1.01,0.05348479747772217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,4,power_law_1.01,0.08786560297012329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,4,power_law_1.01,0.05406079888343811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,4,power_law_1.01,0.10174720287322998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,4,power_law_1.01,0.0546176016330719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,4,power_law_1.01,0.11253119707107544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,4,power_law_1.01,0.05845119953155518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,4,power_law_1.01,0.13916800022125245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,4,power_law_1.01,0.058745598793029784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,4,power_law_1.01,0.17503999471664428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,4,power_law_1.01,0.06330879926681518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,4,power_law_1.01,0.2156287908554077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,4,power_law_1.01,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,4,power_law_1.01,0.277126407623291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,4,power_law_1.01,0.06958079934120179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,4,power_law_1.01,0.4132544040679932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,4,power_law_1.01,0.07723519802093506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,4,power_law_1.01,0.5438144207000732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,4,power_law_1.01,0.08561919927597046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,4,power_law_1.01,0.10179200172424316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,4,power_law_1.01,0.6217472076416015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,4,power_law_1.01,0.11796480417251587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,4,power_law_1.01,0.04510720074176788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,4,power_law_1.01,0.9521599769592285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,4,power_law_1.01,0.15467519760131837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,4,power_law_1.01,0.041484799981117246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,4,power_law_1.01,1.7484479904174806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,4,power_law_1.01,0.19198720455169677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,4,power_law_1.01,0.04472320079803467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,4,power_law_1.01,0.04745599925518036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,4,power_law_1.01,0.26728320121765137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,4,power_law_1.01,0.05225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,4,power_law_1.01,0.3539072036743164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,4,power_law_1.01,0.06222720146179199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,4,power_law_1.01,0.4837503910064697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,4,power_law_1.01,0.06313599944114685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,4,power_law_1.01,0.6083327770233155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,4,power_law_1.01,0.06152960062026978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,4,power_law_1.01,0.8316991806030274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,4,power_law_1.01,0.061440002918243405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,4,power_law_1.01,1.2139904022216796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,4,power_law_1.01,0.06280959844589233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,4,power_law_1.01,0.06260480284690857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,4,power_law_1.01,2.232499122619629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,4,power_law_1.01,0.06470400094985962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,4,power_law_1.01,0.06590080261230469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,4,power_law_1.01,0.06718720197677612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,4,power_law_1.01,0.07203199863433837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,4,power_law_1.01,0.07278079986572265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,4,power_law_1.01,0.07765759825706482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,4,power_law_1.01,0.0883903980255127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,4,power_law_1.01,0.09822720289230347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,4,power_law_1.01,0.1152575969696045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,4,power_law_1.01,0.13894399404525756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,4,power_law_1.01,0.1672320008277893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,4,power_law_1.01,0.21212799549102784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,4,power_law_1.01,0.2993727922439575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,4,power_law_1.01,0.3642879962921143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,4,power_law_1.01,0.016953599452972413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,4,power_law_1.01,0.5525504112243652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,4,power_law_1.01,0.017798399925231932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,4,power_law_1.01,0.6858943939208985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,4,power_law_1.01,0.01852159947156906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,4,power_law_1.01,0.8175168037414551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,4,power_law_1.01,0.019014400243759156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,4,power_law_1.01,1.3262080192565917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,4,power_law_1.01,0.02056960016489029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,4,power_law_1.01,0.022758400440216063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,4,power_law_1.01,2.5186304092407226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,4,power_law_1.01,0.024902400374412537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,4,power_law_1.01,0.024665600061416625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,4,power_law_1.01,0.026163199543952943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,4,power_law_1.01,0.026342400908470155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,4,power_law_1.01,0.026310399174690247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,4,power_law_1.01,0.03351680040359497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,4,power_law_1.01,0.03190400004386902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,4,power_law_1.01,0.03057279884815216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,4,power_law_1.01,0.047839999198913574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,4,power_law_1.01,0.04584319889545441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,4,power_law_1.01,0.0472896009683609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,4,power_law_1.01,0.05432959794998169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,4,power_law_1.01,0.053529602289199826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,4,power_law_1.01,0.05570560097694397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,4,power_law_1.01,0.06133120059967041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,4,power_law_1.01,0.07467520236968994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,4,power_law_1.01,0.08681600093841553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,4,power_law_1.01,0.10817919969558716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,4,power_law_1.01,0.13590400218963622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,4,power_law_1.01,0.181004798412323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,4,power_law_1.01,0.2289664030075073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,4,power_law_1.01,0.28139519691467285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,4,power_law_1.01,0.4169792175292969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,4,power_law_1.01,0.844115161895752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,4,power_law_1.2,0.03768959939479828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,4,power_law_1.2,0.03566080033779144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,4,power_law_1.2,0.03713920116424561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,4,power_law_1.2,0.041152000427246094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,4,power_law_1.2,0.03516800105571747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,4,power_law_1.2,0.04828799962997436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,4,power_law_1.2,0.034457600116729735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,4,power_law_1.2,0.0645632028579712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,4,power_law_1.2,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,4,power_law_1.2,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,4,power_law_1.2,0.039468801021575926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,4,power_law_1.2,0.06604160070419311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,4,power_law_1.2,0.06793599724769592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,4,power_law_1.2,0.04323840141296387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,4,power_law_1.2,0.06734079718589783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,4,power_law_1.2,0.04984320104122162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,4,power_law_1.2,0.06723200082778931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,4,power_law_1.2,0.05084800124168396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,4,power_law_1.2,0.05148800015449524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,4,power_law_1.2,0.06667519807815551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,4,power_law_1.2,0.05343359708786011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,4,power_law_1.2,0.06947839856147767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,4,power_law_1.2,0.054150402545928955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,4,power_law_1.2,0.06883839964866638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,4,power_law_1.2,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,4,power_law_1.2,0.05400320291519165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,4,power_law_1.2,0.056985598802566526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,4,power_law_1.2,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,4,power_law_1.2,0.05624319911003113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,4,power_law_1.2,0.07752959728240967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,4,power_law_1.2,0.060147202014923094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,4,power_law_1.2,0.08269439935684204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,4,power_law_1.2,0.063372802734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,4,power_law_1.2,0.09155200123786926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,4,power_law_1.2,0.06666880249977111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,4,power_law_1.2,0.10536320209503174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,4,power_law_1.2,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,4,power_law_1.2,0.1197119951248169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,4,power_law_1.2,0.08037760257720947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,4,power_law_1.2,0.1462656021118164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,4,power_law_1.2,0.08755840063095092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,4,power_law_1.2,0.1726207971572876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,4,power_law_1.2,0.1018623948097229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,4,power_law_1.2,0.2416383981704712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,4,power_law_1.2,0.12459520101547242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,4,power_law_1.2,0.04529919922351837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,4,power_law_1.2,0.36510720252990725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,4,power_law_1.2,0.16522879600524903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,4,power_law_1.2,0.038848000764846805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,4,power_law_1.2,0.4492095947265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,4,power_law_1.2,0.19460480213165282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,4,power_law_1.2,0.04254080057144165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,4,power_law_1.2,0.5178431987762451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,4,power_law_1.2,0.0474047988653183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,4,power_law_1.2,0.6699327945709228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,4,power_law_1.2,0.05092480182647705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,4,power_law_1.2,1.279520034790039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,4,power_law_1.2,0.0598143994808197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,4,power_law_1.2,2.1136255264282227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,4,power_law_1.2,0.06204800009727478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,4,power_law_1.2,0.063264000415802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,4,power_law_1.2,0.06291840076446534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,4,power_law_1.2,0.062412798404693604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,4,power_law_1.2,0.06299520134925843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,4,power_law_1.2,0.06565120220184326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,4,power_law_1.2,0.06478080153465271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,4,power_law_1.2,0.06779519915580749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,4,power_law_1.2,0.07349119782447815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,4,power_law_1.2,0.07516160011291503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,4,power_law_1.2,0.07876480221748353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,4,power_law_1.2,0.08910719752311706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,4,power_law_1.2,0.09884799718856811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,4,power_law_1.2,0.11975680589675904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,4,power_law_1.2,0.1457856059074402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,4,power_law_1.2,0.19300479888916017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,4,power_law_1.2,0.22060160636901854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,4,power_law_1.2,0.31262080669403075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,4,power_law_1.2,0.3953727960586548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,4,power_law_1.2,0.5926591873168945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,4,power_law_1.2,0.778656005859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,4,power_law_1.2,0.016921600699424742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,4,power_law_1.2,0.01791359931230545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,4,power_law_1.2,1.1186816215515136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,4,power_law_1.2,0.01905920058488846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,4,power_law_1.2,1.8784191131591796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,4,power_law_1.2,0.019840000569820403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,4,power_law_1.2,0.020371200144290925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,4,power_law_1.2,3.255174255371094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,4,power_law_1.2,0.023756800591945647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,4,power_law_1.2,0.025235199928283693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,4,power_law_1.2,0.024537600576877594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,4,power_law_1.2,0.02693760097026825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,4,power_law_1.2,0.026649600267410277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,4,power_law_1.2,0.02654080092906952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,4,power_law_1.2,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,4,power_law_1.2,0.031071999669075014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,4,power_law_1.2,0.030649599432945252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,4,power_law_1.2,0.046463999152183535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,4,power_law_1.2,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,4,power_law_1.2,0.04764800071716309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,4,power_law_1.2,0.05506560206413269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,4,power_law_1.2,0.05280640125274658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,4,power_law_1.2,0.05594879984855652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,4,power_law_1.2,0.06308479905128479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,4,power_law_1.2,0.07697280049324036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,4,power_law_1.2,0.08955519795417785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,4,power_law_1.2,0.11653759479522705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,4,power_law_1.2,0.13964799642562867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,4,power_law_1.2,0.1906559944152832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,4,power_law_1.2,0.2455296039581299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,4,power_law_1.2,0.2959039926528931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,4,power_law_1.2,0.46487040519714357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,4,power_law_1.2,0.880025577545166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,2,balanced,0.036202666660149894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,2,balanced,0.038191998998324074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,2,balanced,0.041093334555625916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,2,balanced,0.050981332858403526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,2,balanced,0.07201066613197327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,2,balanced,0.0735093355178833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,2,balanced,0.07207466661930084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,2,balanced,0.07142400244871776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,2,balanced,0.07282133400440216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,2,balanced,0.07424533367156982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,2,balanced,0.07338133454322815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,2,balanced,0.0772159993648529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,2,balanced,0.07423999905586243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,2,balanced,0.07625066737333934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,2,balanced,0.08183999856313069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,2,balanced,0.08104533453782399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,2,balanced,0.08506666620572408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,4,power_law_1.2,0.2582079887390137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,2,balanced,0.09171733260154724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,4,power_law_1.2,0.338483190536499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,2,balanced,0.09801600376764934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,4,power_law_1.2,0.4996799945831299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,4,power_law_1.2,0.8142720222473144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,2,balanced,0.10995733737945557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,4,power_law_1.2,0.9744447708129883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,4,power_law_1.2,1.6209152221679688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,2,balanced,0.11834667126337688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,4,power_law_1.2,3.0031551361083983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,2,balanced,0.16035200158754984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,2,balanced,0.18123199542363486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,2,balanced,0.03608533243338267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,2,balanced,0.03736533224582672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,2,balanced,0.040949332217375435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,2,balanced,0.2576799988746643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,2,balanced,0.04948266843954722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,2,balanced,0.061717331409454346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,2,balanced,0.06330133477846782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,2,balanced,0.062165334820747375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,2,balanced,0.061119998494784035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,2,balanced,0.06154133379459381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,2,balanced,0.3277920087178548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,2,balanced,0.06407999992370605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,2,balanced,0.06275733311971028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,2,balanced,0.0668693333864212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,2,balanced,0.06513066589832306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,2,balanced,0.06894933183987935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,2,balanced,0.07537066439787547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,2,balanced,0.0748533308506012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,2,balanced,0.07949333389600118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,2,balanced,0.4645173152287801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,2,balanced,0.09092266360918681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,2,balanced,0.10199466347694397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,2,power_law_1.01,0.03591040074825287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,2,balanced,0.11989333232243855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,2,power_law_1.01,0.03697920143604279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,2,power_law_1.01,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,2,power_law_1.01,0.04766719937324524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,2,balanced,0.13749333222707114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,2,power_law_1.01,0.04962559938430786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,2,power_law_1.01,0.06654719710350036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,2,balanced,0.6037866671880087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,2,power_law_1.01,0.06951040029525757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,2,power_law_1.01,0.06936960220336914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,2,balanced,0.18737600247065225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,2,power_law_1.01,0.06941440105438232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,2,power_law_1.01,0.06988800168037415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,2,power_law_1.01,0.07048959732055664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,2,power_law_1.01,0.07309439778327942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,2,power_law_1.01,0.07486079931259156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,2,balanced,0.22393600145975748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,2,power_law_1.01,0.07519999742507935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,2,power_law_1.01,0.08174719810485839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,2,power_law_1.01,0.08133119940757752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,2,power_law_1.01,0.0840831995010376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,2,power_law_1.01,0.09212160110473633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,2,balanced,0.741983970006307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,2,power_law_1.01,0.10117119550704956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,2,balanced,0.31795734167099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,2,power_law_1.01,0.11826560497283936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,2,power_law_1.01,0.13896960020065308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,2,power_law_1.01,0.17498879432678222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,2,power_law_1.01,0.20894079208374022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,2,power_law_1.01,0.28670079708099366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,2,power_law_1.01,0.34697599411010743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,2,power_law_1.01,0.5082943916320801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,2,power_law_1.01,0.6438528060913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,2,balanced,0.41463998953501385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,2,power_law_1.01,0.8252351760864258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,2,power_law_1.01,1.2376704216003418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,2,power_law_1.01,2.302720069885254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,2,balanced,1.1299893061319988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,2,balanced,0.6074133316675822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,2,balanced,0.8014933268229166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,2,power_law_1.01,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,2,power_law_1.01,0.03573119938373566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,2,power_law_1.01,0.03857919871807099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,2,power_law_1.01,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,2,power_law_1.01,0.04653440117835998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,2,power_law_1.01,0.05745279788970947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,2,power_law_1.01,0.0609279990196228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,2,power_law_1.01,0.06047359704971313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,2,power_law_1.01,0.060063999891281125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,2,power_law_1.01,0.06301440000534057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,2,power_law_1.01,0.06442880034446716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,2,power_law_1.01,0.06654719710350036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,2,power_law_1.01,0.06641280055046081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,2,balanced,1.0041600068410237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,2,power_law_1.01,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,2,power_law_1.01,0.07394559979438782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,2,power_law_1.01,0.07745280265808105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,2,power_law_1.01,0.08212479948997498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,2,balanced,2.1829867362976074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,2,power_law_1.01,0.0926912009716034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,2,power_law_1.01,0.10787839889526367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,2,power_law_1.01,0.1307263970375061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,2,power_law_1.01,0.15456000566482545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,2,power_law_1.01,0.19777920246124267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,2,power_law_1.01,0.2526655912399292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,2,power_law_1.01,0.344486403465271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,2,power_law_1.01,0.461740779876709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,2,power_law_1.01,0.6831552028656006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,2,power_law_1.01,0.9125696182250976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,2,power_law_1.01,1.0966976165771485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,2,power_law_1.01,1.7435007095336914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,2,power_law_1.01,3.3719295501708983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,2,balanced,1.5681173006693523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,2,power_law_1.2,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,2,power_law_1.2,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,2,power_law_1.2,0.03988479971885681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,2,power_law_1.2,0.04745599925518036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,2,power_law_1.2,0.0495743989944458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,2,power_law_1.2,0.0666815996170044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,2,power_law_1.2,0.06999679803848266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,2,power_law_1.2,0.0723136007785797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,2,power_law_1.2,0.07056000232696533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,2,power_law_1.2,0.07184000015258789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,2,power_law_1.2,0.07297919988632202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,2,power_law_1.2,0.07554559707641602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,2,power_law_1.2,0.0763584017753601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,2,power_law_1.2,0.07791360020637512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,2,power_law_1.2,0.08081279993057251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,2,power_law_1.2,0.0825215995311737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,2,power_law_1.2,0.08740479946136474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,2,power_law_1.2,0.09313279986381531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,2,power_law_1.2,0.10305919647216796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,2,power_law_1.2,0.11912319660186768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,2,power_law_1.2,0.14120960235595703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,2,power_law_1.2,0.17002240419387818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,2,power_law_1.2,0.20460159778594972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,2,power_law_1.2,0.295033597946167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,2,power_law_1.2,0.3863039970397949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,2,power_law_1.2,0.5315519809722901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,2,power_law_1.2,0.6802752017974854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,2,power_law_1.2,0.8233856201171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,2,power_law_1.2,1.3546175956726074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,2,power_law_1.2,2.7427072525024414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,2,balanced,3.046917279561361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,2,power_law_1.2,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,2,power_law_1.2,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,2,power_law_1.2,0.03797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,2,power_law_1.2,0.0412416011095047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,2,power_law_1.2,0.04474239945411682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,2,power_law_1.2,0.05563520193099976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,2,power_law_1.2,0.060172802209854125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,2,power_law_1.2,0.06017919778823853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,2,power_law_1.2,0.06165120005607605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,2,power_law_1.2,0.06348159909248352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,2,power_law_1.2,0.06309120059013366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,8,balanced,0.04645333190759023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,8,balanced,0.05753066639105479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,2,power_law_1.2,0.06478719711303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,8,balanced,0.07057600220044453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,2,power_law_1.2,0.06702719926834107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,8,balanced,0.10756267110506694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,8,balanced,0.162063995997111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,2,power_law_1.2,0.06915839910507202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,8,balanced,0.15878933668136597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,2,power_law_1.2,0.07638400197029113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,8,balanced,0.16801067193349203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,8,balanced,0.15173332889874777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,2,power_law_1.2,0.07992320060729981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,8,balanced,0.16044800480206808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,2,power_law_1.2,0.08482559919357299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,8,balanced,0.15553067127863565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,8,balanced,0.15409066279729208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,2,power_law_1.2,0.0958079993724823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,8,balanced,0.15877866744995117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,2,power_law_1.2,0.10884480476379395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,8,balanced,0.1548799971739451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,2,power_law_1.2,0.131769597530365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,8,balanced,0.16454399625460306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,2,power_law_1.2,0.15496959686279296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,8,balanced,0.16343999902407327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,2,power_law_1.2,0.21048319339752197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,8,balanced,0.15235732992490134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,2,power_law_1.2,0.2553855895996094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,8,balanced,0.14799466729164124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,2,power_law_1.2,0.3510591983795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,8,balanced,0.15596266587575278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,2,power_law_1.2,0.489247989654541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,8,balanced,0.15360533197720846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,2,power_law_1.2,0.710316801071167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,2,power_law_1.2,0.8810879707336425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,8,balanced,0.16269866625467935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,2,power_law_1.2,1.1988927841186523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,8,balanced,0.17075733343760172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,2,power_law_1.2,1.9808000564575194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,2,power_law_1.2,4.163865661621093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,8,balanced,0.2374133268992106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,8,balanced,0.2362133264541626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,8,balanced,0.3040320078531901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,8,balanced,0.3749653498331706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,8,balanced,0.528165340423584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,8,balanced,0.04206933577855428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,8,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,8,balanced,0.0551093320051829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,8,balanced,0.07334400216738383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,8,balanced,0.11582400401433308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,8,balanced,0.04033066580692927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,8,balanced,0.11046399672826131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,8,balanced,0.677344004313151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,8,balanced,0.046053335070610046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,8,balanced,0.10406399766604106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,8,balanced,0.05714133381843567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,8,balanced,0.10473066568374634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,8,balanced,0.07036266724268596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,8,balanced,0.10315733154614766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,8,balanced,0.11011733611424764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,8,balanced,0.1067519982655843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,8,balanced,0.1002346674601237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,8,balanced,0.09910933176676433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,8,balanced,0.10251200199127197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,8,balanced,0.1030613382657369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,8,balanced,0.1027733286221822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,8,balanced,0.11018133163452148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,8,balanced,0.09842666983604431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,8,balanced,0.0983733336130778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,8,balanced,0.11053333679835002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,8,balanced,0.09821333487828572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,8,balanced,0.11226133505503337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,8,balanced,0.1109920044740041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,8,balanced,0.10522666573524475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,8,balanced,0.10794132947921753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,8,balanced,0.10385066270828247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,8,balanced,0.09944533308347066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,8,balanced,0.8330667018890381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,8,balanced,0.10386133193969727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,8,balanced,0.10989866654078166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,8,balanced,0.10673600435256958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,8,balanced,0.11227200428644817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,8,balanced,0.1104906698067983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,8,balanced,0.11302933096885681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,8,balanced,0.11310399572054546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,8,balanced,0.12526933352152506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,8,balanced,0.12038933237393697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,8,balanced,0.13205333550771078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,8,balanced,0.18279999494552612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,8,balanced,0.14427199959754944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,8,balanced,0.19181332985560098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,8,balanced,0.18798933426539102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,8,balanced,0.25408534208933514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,8,balanced,0.2104533314704895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,8,balanced,1.27729598681132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,8,balanced,0.31063999732335407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,8,balanced,0.2837653358777364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,8,balanced,0.3537600040435791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,8,balanced,0.43192001183827716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,8,balanced,0.5542879899342855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,8,balanced,0.5008533398310343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,8,balanced,0.023413332800070446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,8,balanced,0.02622399975856145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,8,balanced,0.028197333216667175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,8,balanced,0.03272533416748047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,8,balanced,0.0488319993019104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,8,balanced,0.05085866649945577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,8,balanced,0.06425066788991292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,8,balanced,0.060975998640060425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,8,balanced,0.09177066882451375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,8,balanced,0.08922666311264038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,8,balanced,0.08044800162315369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,8,balanced,0.09549333651860555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,8,balanced,0.680298646291097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,8,balanced,2.5140105883280435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,8,balanced,0.09609599908192952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,8,balanced,0.6455413500467936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,8,balanced,0.09644266963005066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,8,balanced,0.1234773298104604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,8,balanced,0.11968533198038737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,8,balanced,0.10842133561770122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,8,balanced,0.11724266409873962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,8,balanced,0.1309706668059031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,8,balanced,0.14366933703422546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,8,balanced,0.1587999959786733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,8,balanced,0.18765334288279215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,8,balanced,1.03438401222229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,8,balanced,0.7929333051045736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,8,balanced,0.21031999588012695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,8,balanced,0.3296053409576416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,8,balanced,0.38118934631347656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,8,balanced,0.5543306668599447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,8,balanced,1.2172959645589192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,8,balanced,0.7265439828236898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,8,balanced,2.022058645884196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,8,balanced,0.905839999516805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,8,balanced,1.4297919273376465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,8,power_law_1.01,0.09800320267677307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,8,power_law_1.01,0.0809664011001587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,8,power_law_1.01,0.07164160013198853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,8,power_law_1.01,0.10884480476379395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,8,power_law_1.01,0.10908160209655762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,8,power_law_1.01,0.14703999757766723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,8,balanced,2.3723626136779785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,8,power_law_1.01,0.1542400002479553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,8,power_law_1.01,0.15152640342712403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,8,power_law_1.01,0.1491520047187805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,8,power_law_1.01,0.15224319696426392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,8,power_law_1.01,0.15550719499588012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,8,power_law_1.01,0.15749119520187377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,8,power_law_1.01,0.15652480125427246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,8,power_law_1.01,0.150873601436615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,8,power_law_1.01,0.15791360139846802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,8,power_law_1.01,0.16257920265197753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,8,power_law_1.01,0.17331199645996093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,8,power_law_1.01,0.19028480052948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,8,power_law_1.01,0.20558080673217774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,8,power_law_1.01,0.24444160461425782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,8,power_law_1.01,0.26169600486755373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,8,power_law_1.01,0.3201472043991089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,8,power_law_1.01,0.39967360496521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,8,power_law_1.01,0.5139008045196534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,8,power_law_1.01,0.7348608016967774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,8,power_law_1.01,0.8922944068908691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,8,power_law_1.01,1.2800127983093261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,8,power_law_1.01,1.485324764251709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,8,power_law_1.01,1.9925376892089843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,8,power_law_1.01,3.9499454498291016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,8,balanced,2.8227307001749673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,8,power_law_1.01,0.07742720246315002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,8,power_law_1.01,0.054841601848602296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,8,power_law_1.01,0.05612800121307373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,8,power_law_1.01,0.07134720087051391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,8,power_law_1.01,0.07859200239181519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,8,power_law_1.01,0.09441279768943786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,8,power_law_1.01,0.09859840273857116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,8,power_law_1.01,0.1019711971282959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,8,power_law_1.01,0.10088319778442383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,8,power_law_1.01,0.09884160161018371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,8,power_law_1.01,0.10526080131530761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,8,power_law_1.01,0.10190080404281616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,8,power_law_1.01,0.10621440410614014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,8,power_law_1.01,0.10205440521240235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,8,power_law_1.01,0.10841599702835084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,8,power_law_1.01,0.10751999616622925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,8,power_law_1.01,0.11857919692993164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,8,power_law_1.01,0.1347584009170532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,8,power_law_1.01,0.14603519439697266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,8,power_law_1.01,0.1672127962112427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,8,power_law_1.01,0.1972864031791687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,8,power_law_1.01,0.25258240699768064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,8,power_law_1.01,0.32187519073486326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,8,power_law_1.01,0.41030402183532716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,8,power_law_1.01,0.4952064037322998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,8,power_law_1.01,0.7495168209075928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,8,power_law_1.01,1.0124799728393554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,8,power_law_1.01,1.1700991630554198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,8,power_law_1.01,2.1593856811523438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,8,power_law_1.01,3.4699775695800783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,8,power_law_1.01,0.07525119781494141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,8,power_law_1.01,0.06369919776916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,8,power_law_1.01,0.05491200089454651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,8,power_law_1.01,0.0757311999797821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,8,power_law_1.01,0.07683839797973632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,8,power_law_1.01,0.10116480588912964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,8,power_law_1.01,0.09738240242004395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,8,power_law_1.01,0.10149120092391968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,8,power_law_1.01,0.09750400185585022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,8,power_law_1.01,0.10632319450378418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,8,power_law_1.01,0.10508160591125489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,8,power_law_1.01,0.1071936011314392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,8,power_law_1.01,0.1034559965133667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,8,power_law_1.01,0.10610560178756714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,8,power_law_1.01,0.11324800252914428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,8,power_law_1.01,0.11272319555282592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,8,power_law_1.01,0.11832319498062134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,8,power_law_1.01,0.132697594165802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.01,0.13983999490737914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.01,0.17354880571365355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.01,0.18919039964675904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.01,0.23168001174926758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.01,0.2982656002044678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.01,0.38081281185150145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,8,power_law_1.01,0.02648960053920746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,8,power_law_1.01,0.024153600633144378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.01,0.48336639404296877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,8,power_law_1.01,0.02558720111846924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.01,0.6657408237457275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,8,power_law_1.01,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.01,0.8487232208251954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,8,power_law_1.01,0.03614720106124878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,8,power_law_1.01,1.0402688026428222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,8,power_law_1.01,0.04505600035190582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,8,power_law_1.01,1.7560895919799804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,8,power_law_1.01,0.05541120171546936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,8,power_law_1.01,0.050758397579193114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,8,power_law_1.01,3.1689599990844726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,8,power_law_1.01,0.07423359751701356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,8,power_law_1.01,0.07272319793701172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,8,power_law_1.01,0.06403840184211732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,8,power_law_1.01,0.08597760200500489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,8,power_law_1.2,0.09631999731063842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,8,power_law_1.01,0.08599680066108703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,8,power_law_1.2,0.06038399934768677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,8,power_law_1.01,0.0836736023426056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,8,power_law_1.2,0.06563199758529663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,8,power_law_1.01,0.10103039741516114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,8,power_law_1.2,0.08586879968643188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,8,power_law_1.01,0.10117759704589843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,8,power_law_1.2,0.10936319828033447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,8,power_law_1.01,0.09120000004768372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,8,power_law_1.2,0.14455679655075074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,8,power_law_1.01,0.0890496015548706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,8,power_law_1.2,0.1495296001434326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.01,0.08880000114440918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,8,power_law_1.2,0.15310720205307007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.01,0.10069760084152221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,8,power_law_1.2,0.1565951943397522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.01,0.1139456033706665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,8,power_law_1.2,0.15827200412750245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.01,0.13686399459838866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,8,power_law_1.2,0.1601855993270874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.01,0.16906239986419677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,8,power_law_1.2,0.15703680515289306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.01,0.22519679069519044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,8,power_law_1.2,0.1518336057662964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.01,0.2847743988037109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,8,power_law_1.2,0.14776959419250488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.01,0.42540798187255857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,8,power_law_1.2,0.16852480173110962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.01,0.5419519901275635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,8,power_law_1.2,0.17179520130157472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,8,power_law_1.01,0.6494143962860107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,8,power_law_1.2,0.17004799842834473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,8,power_law_1.01,0.9962623596191407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,8,power_law_1.2,0.1941823959350586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,8,power_law_1.2,0.2176896095275879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,8,power_law_1.01,2.0146303176879883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,8,power_law_1.2,0.24030721187591553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,8,power_law_1.2,0.2859776020050049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,8,power_law_1.2,0.3736511945724487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,8,power_law_1.2,0.467955207824707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,8,power_law_1.2,0.5597055912017822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,8,power_law_1.2,0.7717760086059571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,8,power_law_1.2,1.1333120346069336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,8,power_law_1.2,1.5061759948730469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,8,power_law_1.2,2.011539268493652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,8,power_law_1.2,3.1500736236572267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,8,power_law_1.2,7.662828826904297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,8,power_law_1.2,0.0759872019290924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,8,power_law_1.2,0.0527616024017334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,8,power_law_1.2,0.053337597846984865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,8,power_law_1.2,0.06562560200691223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,8,power_law_1.2,0.07928959727287292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,8,power_law_1.2,0.09591680169105529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,8,power_law_1.2,0.10108799934387207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,8,power_law_1.2,0.10471680164337158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,8,power_law_1.2,0.09982720017433167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,8,power_law_1.2,0.10116480588912964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,8,power_law_1.2,0.1044927954673767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,8,power_law_1.2,0.10283520221710205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,8,power_law_1.2,0.1004032015800476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,8,power_law_1.2,0.10645760297775268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,8,power_law_1.2,0.11118079423904419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,8,power_law_1.2,0.11868159770965576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,8,power_law_1.2,0.12241920232772827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,8,power_law_1.2,0.13428479433059692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,8,power_law_1.2,0.15857280492782594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,8,power_law_1.2,0.17738879919052125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,8,power_law_1.2,0.2170624017715454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,8,power_law_1.2,0.2725951910018921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,8,power_law_1.2,0.3537343978881836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,8,power_law_1.2,0.45418238639831543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,8,power_law_1.2,0.5935167789459228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,8,power_law_1.2,0.9031999588012696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,8,power_law_1.2,1.305292797088623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,8,power_law_1.2,1.7035392761230468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,8,power_law_1.2,2.3483135223388674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,8,power_law_1.2,5.637030410766601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,8,power_law_1.2,0.07919999957084656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,8,power_law_1.2,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,8,power_law_1.2,0.05925120115280151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,8,power_law_1.2,0.0743615984916687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,8,power_law_1.2,0.08090239763259888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,8,power_law_1.2,0.10301439762115479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,8,power_law_1.2,0.1003775954246521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,8,power_law_1.2,0.10408960580825806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,8,power_law_1.2,0.10744320154190064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,8,power_law_1.2,0.10071040391921997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,8,power_law_1.2,0.10849920511245728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,8,power_law_1.2,0.10592639446258545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,8,power_law_1.2,0.1087231993675232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,8,power_law_1.2,0.10817919969558716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,8,power_law_1.2,0.11301120519638061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,8,power_law_1.2,0.11921279430389405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,8,power_law_1.2,0.1243008017539978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,8,power_law_1.2,0.1323456048965454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.2,0.15733760595321655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.2,0.18431999683380126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.2,0.19438719749450684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.2,0.2812544107437134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.2,0.30380799770355227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.2,0.48295040130615235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.2,0.523308801651001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.2,0.8276736259460449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,8,power_law_1.2,0.02630400061607361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.2,1.1049728393554688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,8,power_law_1.2,0.022009600698947907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,8,power_law_1.2,0.026342400908470155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,8,power_law_1.2,1.3370944023132325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,8,power_law_1.2,0.028332799673080444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,8,power_law_1.2,2.163942337036133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,8,power_law_1.2,0.031744000315666196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,8,power_law_1.2,0.04628480076789856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,8,power_law_1.2,4.460992050170899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,8,power_law_1.2,0.05348479747772217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,8,power_law_1.2,0.05198079943656921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,8,power_law_1.2,0.07438719868659974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,8,power_law_1.2,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,8,power_law_1.2,0.06845440268516541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,8,power_law_1.2,0.08208640217781067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,8,power_law_1.2,0.08492159843444824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,4,balanced,0.04231466849644979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,8,power_law_1.2,0.08309760093688964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,4,balanced,0.05463466544946035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,4,balanced,0.06977599859237671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,8,power_law_1.2,0.09768959879875183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,4,balanced,0.10752532879511516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,8,power_law_1.2,0.09349759817123413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,4,balanced,0.15940800309181213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,8,power_law_1.2,0.09390079975128174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,4,balanced,0.157258669535319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,4,balanced,0.15761066476504007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,8,power_law_1.2,0.09464960098266602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,4,balanced,0.14459199706713358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.2,0.10055680274963379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,4,balanced,0.14991999665896097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.2,0.11085439920425415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,4,balanced,0.14969066778818765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.2,0.11594879627227783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,4,balanced,0.14085867007573447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.2,0.1522879958152771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,4,balanced,0.15722666184107462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,4,balanced,0.14496533075968424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.2,0.18108160495758058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,4,balanced,0.14443733294804892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.2,0.23644800186157228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,4,balanced,0.15026666720708212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.2,0.305516791343689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,4,balanced,0.14994133512179056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.2,0.4507008075714111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,4,balanced,0.15476800004641214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.2,0.5959936141967773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,8,power_law_1.2,0.7536384105682373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,4,balanced,0.15819733341534933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,8,power_law_1.2,1.2126144409179687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,4,balanced,0.16082666317621866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,8,power_law_1.2,2.3572927474975587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,4,balanced,0.1758240063985189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,4,balanced,0.18800000349680582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,4,balanced,0.2582239905993144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,4,balanced,0.27452800671259564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,4,balanced,0.35392534732818604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,4,balanced,0.4453493356704712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,4,balanced,0.040005333721637726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,4,balanced,0.04318933188915253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,4,balanced,0.05136533578236898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,4,balanced,0.07425066828727722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,4,balanced,0.10295466581980388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,4,balanced,0.10594666997591655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,4,balanced,0.10197866956392924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,4,balanced,0.6298133134841919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,4,balanced,0.10204799969991048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,4,balanced,0.09961066643397014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,4,balanced,0.10019733508427937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,4,balanced,0.09819199641545613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,4,balanced,0.10514133175214131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,4,balanced,0.09857066472371419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,4,balanced,0.10037333766619365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,4,balanced,0.11117333173751831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,4,balanced,0.10782399773597717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,4,balanced,0.11455466349919637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,4,balanced,0.816543976465861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,4,balanced,0.12247467041015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,4,balanced,0.13006400068600973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,4,balanced,0.1474399964014689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,4,balanced,0.16328533490498862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,4,balanced,0.22098666429519653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,4,balanced,1.0007466475168865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,4,balanced,0.2577066620190938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,4,balanced,0.3547146717707316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,4,balanced,0.045509333411852516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,4,balanced,0.4464319944381714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,4,balanced,0.049285332361857094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,4,balanced,0.05653333167235056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,4,balanced,0.08212799827257793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,4,balanced,0.11512000362078349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,4,balanced,0.11456533273061116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,4,balanced,0.10842667023340861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,4,balanced,0.1113759974638621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,4,balanced,0.10693867007891338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,4,balanced,0.10544000069300334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,4,balanced,0.1053546667098999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,4,balanced,0.6402080059051514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,4,balanced,0.1120746632417043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,4,balanced,1.549909273783366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,4,balanced,0.10842133561770122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,4,balanced,0.10126933455467224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,4,balanced,0.11986666917800903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,4,balanced,0.11531733473141988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,4,balanced,0.1107413371404012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,4,balanced,0.11864533027013142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,4,balanced,0.12730133533477783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,4,balanced,0.831109364827474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,4,balanced,0.137472003698349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,4,balanced,0.02164799968401591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,4,balanced,0.15308266878128052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,4,balanced,0.023370665808518726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,4,balanced,0.02608533451954524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,4,balanced,0.03052799900372823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,4,balanced,0.04814933240413666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,4,balanced,0.05041066805521647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,4,balanced,0.21945067246754965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,4,balanced,0.05266666909058889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,4,balanced,0.05485333502292633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,4,balanced,0.06387199958165486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,4,balanced,0.06121600170930227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,4,balanced,0.061520000298817955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,4,balanced,0.2444159984588623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,4,balanced,1.0254080295562744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,4,balanced,0.08879466851552327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,4,balanced,0.08306666711966197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,4,balanced,0.08530666430791219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,4,balanced,0.09965866804122925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,4,balanced,0.10139200091362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,4,balanced,0.3343573411305745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,4,balanced,0.10673066973686218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,4,balanced,0.11769066254297893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,4,balanced,0.12294399738311768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,4,balanced,0.13547733426094055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,4,balanced,0.14655466874440512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,4,balanced,0.41790934403737384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,4,balanced,0.1613920032978058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,4,balanced,3.059274673461914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,4,balanced,0.16924800475438437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,4,balanced,1.5916372934977214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,4,balanced,0.27155200640360516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,4,balanced,0.5951360066731771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,4,balanced,0.29734400908152264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,4,balanced,0.4225120147069295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,4,balanced,0.7665013472239176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,4,balanced,0.5510773261388143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,4,balanced,0.9366559982299805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,4,balanced,3.0991945266723633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,4,balanced,0.6812960306803385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,4,balanced,1.0868159929911296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,4,balanced,1.4337280591328938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,4,power_law_1.01,0.06401919722557067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,4,power_law_1.01,0.05496960282325745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,4,power_law_1.01,0.06417919993400574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,4,power_law_1.01,0.09237120151519776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,4,power_law_1.01,0.10334080457687378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,4,power_law_1.01,0.14505599737167357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,4,power_law_1.01,0.1401087999343872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,4,power_law_1.01,0.14197759628295897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,4,balanced,2.137925306955973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,4,power_law_1.01,0.14472960233688353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,4,power_law_1.01,0.14498560428619384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,4,power_law_1.01,0.15251200199127196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,4,power_law_1.01,0.15091840028762818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,4,power_law_1.01,0.14991359710693358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,4,power_law_1.01,0.15422719717025757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,4,power_law_1.01,0.15198080539703368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,4,power_law_1.01,0.1554111957550049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,4,power_law_1.01,0.16560640335083007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,4,power_law_1.01,0.1878592014312744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,4,power_law_1.01,0.19141759872436523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,4,power_law_1.01,0.22420480251312255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,4,balanced,2.758234659830729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,4,power_law_1.01,0.2640127897262573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,4,power_law_1.01,0.3406847953796387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,4,power_law_1.01,0.38461439609527587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,4,power_law_1.01,0.49176959991455077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,4,power_law_1.01,0.6527232170104981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,4,power_law_1.01,0.8381952285766602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,4,power_law_1.01,1.185536003112793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,4,power_law_1.01,1.3522111892700195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,4,power_law_1.01,2.1409088134765626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,4,power_law_1.01,3.9074111938476563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,4,power_law_1.01,0.05449600219726562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,4,power_law_1.01,0.04523519873619079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,4,power_law_1.01,0.05129600167274475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,4,power_law_1.01,0.0595583975315094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,4,power_law_1.01,0.07420160174369812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,4,power_law_1.01,0.09205759763717651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,4,power_law_1.01,0.09516159892082214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,4,power_law_1.01,0.09610880017280579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,4,power_law_1.01,0.09859200119972229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,4,power_law_1.01,0.0976639986038208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,4,power_law_1.01,0.10044159889221191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,4,power_law_1.01,0.1027008056640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,4,power_law_1.01,0.10588799715042115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,4,power_law_1.01,0.1076159954071045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,4,power_law_1.01,0.11119999885559081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,4,power_law_1.01,0.11414400339126587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,4,power_law_1.01,0.11904000043869019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,4,power_law_1.01,0.13639039993286134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,4,power_law_1.01,0.14885760545730592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,4,power_law_1.01,0.17795200347900392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,4,power_law_1.01,0.20329599380493163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,4,power_law_1.01,0.2688512086868286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,4,power_law_1.01,0.33501439094543456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,4,power_law_1.01,0.4651199817657471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,4,power_law_1.01,0.575596809387207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,4,power_law_1.01,0.7677184104919433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,4,power_law_1.01,1.0855487823486327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,4,power_law_1.01,1.3124287605285645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,4,power_law_1.01,2.024710464477539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,4,power_law_1.01,3.877267074584961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,4,power_law_1.01,0.06240000128746033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,4,power_law_1.01,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,4,power_law_1.01,0.056415998935699464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,4,power_law_1.01,0.07044479846954346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,4,power_law_1.01,0.08104959726333619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,4,power_law_1.01,0.10101120471954346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,4,power_law_1.01,0.1081279993057251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,4,power_law_1.01,0.10472960472106933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,4,power_law_1.01,0.1063040018081665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,4,power_law_1.01,0.1056831955909729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,4,power_law_1.01,0.10718079805374145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,4,power_law_1.01,0.10762879848480225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,4,power_law_1.01,0.1108672022819519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,4,power_law_1.01,0.11153279542922974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,4,power_law_1.01,0.021849599480628968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,4,power_law_1.01,0.11475839614868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,4,power_law_1.01,0.02210559993982315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,4,power_law_1.01,0.11584000587463379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,4,power_law_1.01,0.024665600061416625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,4,power_law_1.01,0.12253439426422119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,4,power_law_1.01,0.02928000092506409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,4,power_law_1.01,0.13316479921340943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,4,power_law_1.01,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.01,0.15837440490722657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,4,power_law_1.01,0.04179840087890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.01,0.17821439504623413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,4,power_law_1.01,0.04743039906024933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.01,0.2001408100128174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,4,power_law_1.2,0.06538879871368408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,4,power_law_1.01,0.04596480131149292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.01,0.2828991889953613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,4,power_law_1.2,0.05144960284233093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,4,power_law_1.01,0.04968959987163544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.01,0.3168191909790039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,4,power_law_1.2,0.06359040141105651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,4,power_law_1.01,0.0491456001996994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.01,0.42490239143371583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,4,power_law_1.2,0.08143360018730164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,4,power_law_1.01,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.01,0.556876802444458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,4,power_law_1.2,0.09564160108566284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,4,power_law_1.01,0.06629760265350342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.01,0.7598720073699952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,4,power_law_1.2,0.1425279974937439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,4,power_law_1.01,0.05761920213699341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.01,1.0186304092407226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,4,power_law_1.2,0.14265600442886353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,4,power_law_1.01,0.0597055971622467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,4,power_law_1.01,1.1676799774169921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,4,power_law_1.2,0.1447872042655945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,4,power_law_1.01,0.08558719754219055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,4,power_law_1.01,1.7557952880859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,4,power_law_1.2,0.14821759462356568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,4,power_law_1.01,0.08408960103988647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,4,power_law_1.2,0.1489408016204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,4,power_law_1.01,0.09275519847869873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,4,power_law_1.01,3.822188949584961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,4,power_law_1.2,0.15340800285339357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,4,power_law_1.01,0.09652479887008666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,4,power_law_1.2,0.15472639799118043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.01,0.08938239812850952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,4,power_law_1.2,0.14802559614181518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.01,0.10054399967193603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,4,power_law_1.2,0.15237120389938355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.01,0.10975359678268433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,4,power_law_1.2,0.15642240047454833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.01,0.13292800188064574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,4,power_law_1.2,0.16367360353469848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.01,0.15146880149841307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,4,power_law_1.2,0.1706112027168274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.01,0.1895807981491089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,4,power_law_1.2,0.1883263945579529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.01,0.24746239185333252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,4,power_law_1.2,0.20042240619659424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.01,0.3549247980117798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,4,power_law_1.2,0.2307904005050659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.01,0.44715518951416017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,4,power_law_1.2,0.2898175954818726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,4,power_law_1.01,0.555628776550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,4,power_law_1.2,0.34618239402770995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,4,power_law_1.01,0.8399807929992675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,4,power_law_1.2,0.3893503904342651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,4,power_law_1.2,0.5497663974761963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,4,power_law_1.01,1.6675775527954102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,4,power_law_1.2,0.7483200073242188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,4,power_law_1.2,0.9766079902648925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,4,power_law_1.2,1.3699392318725585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,4,power_law_1.2,1.5580479621887207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,4,power_law_1.2,2.7890047073364257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,4,power_law_1.2,4.970835113525391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,4,power_law_1.2,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,4,power_law_1.2,0.04455040097236633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,4,power_law_1.2,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,4,power_law_1.2,0.06044160127639771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,4,power_law_1.2,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,4,power_law_1.2,0.09237120151519776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,4,power_law_1.2,0.09423360228538513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,4,power_law_1.2,0.09751679897308349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,4,power_law_1.2,0.10005760192871094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,4,power_law_1.2,0.09765120148658753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,4,power_law_1.2,0.10396159887313842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,4,power_law_1.2,0.10523519515991211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,4,power_law_1.2,0.10539519786834717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,4,power_law_1.2,0.1093951940536499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,4,power_law_1.2,0.11276160478591919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,4,power_law_1.2,0.1155519962310791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,4,power_law_1.2,0.11797120571136474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,4,power_law_1.2,0.14035199880599974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,4,power_law_1.2,0.15470720529556276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,4,power_law_1.2,0.18048640489578247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,4,power_law_1.2,0.21330559253692627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,4,power_law_1.2,0.2818687915802002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,4,power_law_1.2,0.3389120101928711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,4,power_law_1.2,0.49764480590820315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,4,power_law_1.2,0.6552576065063477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,4,power_law_1.2,0.8934528350830078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,4,power_law_1.2,1.255782413482666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,4,power_law_1.2,1.4809023857116699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,4,power_law_1.2,2.416486358642578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,4,power_law_1.2,5.125331115722656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,4,power_law_1.2,0.05898879766464234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,4,power_law_1.2,0.04508799910545349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,4,power_law_1.2,0.051532799005508424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,4,power_law_1.2,0.06389120221138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,4,power_law_1.2,0.07622399926185608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,4,power_law_1.2,0.10592639446258545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,4,power_law_1.2,0.10646400451660157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,4,power_law_1.2,0.10311679840087891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,4,power_law_1.2,0.11028480529785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,4,power_law_1.2,0.1087488055229187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,4,power_law_1.2,0.11148799657821655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,4,power_law_1.2,0.11234560012817382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,4,power_law_1.2,0.11059199571609497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,4,power_law_1.2,0.11427199840545654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,4,power_law_1.2,0.11755520105361938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,4,power_law_1.2,0.12472959756851196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,4,power_law_1.2,0.12867200374603271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,4,power_law_1.2,0.1477311968803406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.2,0.15875200033187867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.2,0.18626559972763063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.2,0.20145280361175538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.2,0.2850559949874878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.2,0.33975679874420167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,4,power_law_1.2,0.021721599996089934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.2,0.5057024002075196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,4,power_law_1.2,0.02056960016489029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.2,0.541811180114746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,4,power_law_1.2,0.02454400062561035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,4,power_law_1.2,0.026604801416397095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.2,0.8494336128234863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,4,power_law_1.2,0.03255679905414581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.2,1.1980352401733398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,4,power_law_1.2,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,4,power_law_1.2,1.4142784118652343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,4,power_law_1.2,0.0465472012758255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,4,power_law_1.2,2.406329536437988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,4,power_law_1.2,0.04731520116329193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,4,power_law_1.2,0.052095997333526614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,2,balanced,0.04605866471926371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,4,power_law_1.2,5.030937576293946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,2,balanced,0.05264000097910563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,4,power_law_1.2,0.04997119903564453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,2,balanced,0.06954666475454967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,4,power_law_1.2,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,2,balanced,0.10507200161616008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,4,power_law_1.2,0.06368640065193176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,2,balanced,0.15222400426864624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,4,power_law_1.2,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,2,balanced,0.15110400319099426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,2,balanced,0.1474133332570394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,4,power_law_1.2,0.057183998823165896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,2,balanced,0.14563199877738953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,4,power_law_1.2,0.0847487986087799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,2,balanced,0.151936004559199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,4,power_law_1.2,0.08798720240592957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,2,balanced,0.14683733383814493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,4,power_law_1.2,0.09950079917907714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,2,balanced,0.14727466305096945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,4,power_law_1.2,0.09453439712524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,2,balanced,0.15018133322397867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.2,0.09208959937095643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,2,balanced,0.15050666530927023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.2,0.1016767978668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,2,balanced,0.1504693329334259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.2,0.11637760400772094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,2,balanced,0.15737600127855936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.2,0.13372160196304322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,2,balanced,0.1585919956366221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.2,0.1589951992034912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,2,balanced,0.16364799936612448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.2,0.20849919319152832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,2,balanced,0.17549866437911987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.2,0.26881918907165525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.2,0.3736448049545288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,2,balanced,0.18508267402648926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.2,0.4725503921508789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,4,power_law_1.2,0.6053247928619385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,2,balanced,0.2088586688041687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,4,power_law_1.2,0.8995200157165527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,2,balanced,0.2284160057703654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,4,power_law_1.2,1.7484096527099608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,2,balanced,0.31031999985376996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,2,balanced,0.0421973317861557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,2,balanced,0.046256000796953835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,2,balanced,0.05288533369700114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,2,balanced,0.3555306593577067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,2,balanced,0.08019199967384338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,2,balanced,0.10949333508809407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,2,balanced,0.10886399944623311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,2,balanced,0.10714133580525716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,2,balanced,0.10833600163459778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,2,balanced,0.10885866483052571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,2,balanced,0.11034666498502095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,2,balanced,0.4757973353068034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,2,balanced,0.10796800255775452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,2,balanced,0.11573867003122966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,2,balanced,0.11363733808199565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,2,balanced,0.11599999666213989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,2,balanced,0.12379733721415202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,2,balanced,0.12478400270144145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,2,balanced,0.13052800297737122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,2,balanced,0.14665599664052328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,2,balanced,0.6118613481521606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,2,balanced,0.16149333119392395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,2,balanced,0.18502932786941528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,2,balanced,0.21058666706085205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,2,balanced,0.29581334193547565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,2,balanced,0.8654080231984457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,2,balanced,0.353274663289388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,2,balanced,0.4947733481725057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,2,balanced,0.63865065574646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,2,balanced,1.1173866589864094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,2,balanced,0.9214026927947998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,2,balanced,1.3691306114196777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,2,balanced,1.1942026615142822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,2,balanced,0.050570666790008545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,2,balanced,0.055205335219701133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,2,balanced,0.07150400181611379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,2,balanced,0.09499200185139973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,2,balanced,0.14138666788736978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,2,balanced,0.13287466764450073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,2,balanced,0.12688000003496805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,2,balanced,0.12760532895723978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,2,balanced,0.12963199615478516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,2,balanced,0.12703466415405273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,2,balanced,0.1285599966843923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,2,balanced,0.1302293340365092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,2,balanced,0.13314666350682577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,2,balanced,0.12867732842763266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,2,balanced,1.4796160062154133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,2,balanced,0.1418880025545756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,2,balanced,0.13944000005722046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,2,balanced,0.14313066999117532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,2,balanced,0.15930133064587912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,2,balanced,0.17298134167989096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,2,balanced,0.20334933201471964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,2,balanced,0.22619734207789102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,2,balanced,2.139845371246338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,2,balanced,0.02184533327817917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,2,balanced,0.3397386471430461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,2,balanced,0.023775999744733173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,2,balanced,0.02446399877468745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,2,balanced,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,2,balanced,0.0491946687301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,2,balanced,0.050997331738471985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,2,balanced,0.39554135004679364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,2,balanced,0.05203733344872793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,2,balanced,0.05358933409055074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,2,balanced,2.3047146797180176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,2,balanced,0.05658133327960968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,2,balanced,0.05580799778302511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,2,balanced,0.05845866600672404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,2,balanced,0.5536373456319174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,2,balanced,0.06314133107662201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,2,balanced,0.0652159998814265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,2,balanced,0.0649599979321162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,2,balanced,0.08339200417200725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,2,balanced,0.08143466711044312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,2,balanced,0.703221321105957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,2,balanced,0.08163733283678691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,2,balanced,0.11661332845687866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,2,balanced,0.12227200468381245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,2,balanced,0.14521066347757974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,2,balanced,0.15213333566983542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,2,balanced,0.9993226528167725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,2,balanced,0.15917866428693137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,2,balanced,0.1691946585973104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,2,balanced,1.3044853210449219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,2,balanced,0.27663999795913696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,2,balanced,0.291920006275177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,2,balanced,4.517146746317546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,2,balanced,4.196991920471191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,2,balanced,1.6026293436686199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,2,balanced,0.4141226609547933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,2,balanced,0.53984534740448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,2,balanced,2.505173365275065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,2,balanced,0.6657919883728027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,2,balanced,1.063706636428833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,2,balanced,4.879941304524739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,2,power_law_1.01,0.05275520086288452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,2,power_law_1.01,0.05549439787864685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,2,power_law_1.01,0.061273598670959474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,2,power_law_1.01,0.07838079929351807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,2,power_law_1.01,0.10136959552764893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,2,power_law_1.01,0.14350080490112305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,2,power_law_1.01,0.14355839490890504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,2,power_law_1.01,0.1418944001197815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,2,power_law_1.01,0.14746880531311035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,2,balanced,2.0868159929911294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,2,power_law_1.01,0.14562560319900514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,2,power_law_1.01,0.15014400482177734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,2,power_law_1.01,0.04469760060310364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,2,power_law_1.01,0.1513856053352356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,2,power_law_1.01,0.0455808013677597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,2,power_law_1.01,0.1565119981765747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,2,power_law_1.01,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,2,power_law_1.01,0.15459200143814086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,2,power_law_1.01,0.06435199975967407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,2,power_law_1.01,0.16213760375976563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,2,power_law_1.01,0.07750399708747864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,2,power_law_1.01,0.1619968056678772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,2,power_law_1.01,0.10060160160064698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,2,power_law_1.01,0.16900479793548584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,2,power_law_1.01,0.10275839567184449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,2,power_law_1.01,0.10526080131530761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,2,power_law_1.01,0.1824447989463806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,2,power_law_1.01,0.1082111954689026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,2,power_law_1.01,0.20402560234069825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,2,power_law_1.01,0.10680320262908935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,2,power_law_1.01,0.2334144115447998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,2,power_law_1.01,0.1087615966796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,2,power_law_1.01,0.28158719539642335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,2,power_law_1.01,0.11184639930725097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,2,power_law_1.01,0.3499840021133423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,2,power_law_1.01,0.11392639875411988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,2,power_law_1.01,0.41074562072753906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,2,power_law_1.01,0.11704319715499878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,2,power_law_1.01,0.5550528049468995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,2,power_law_1.01,0.12446719408035278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,2,power_law_1.01,0.6886464118957519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,2,power_law_1.01,0.12762240171432496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,2,power_law_1.01,1.0103360176086427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,2,power_law_1.01,0.13598079681396485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,2,power_law_1.01,1.2985216140747071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,2,power_law_1.01,0.15627520084381102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,2,power_law_1.01,1.547916793823242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,2,power_law_1.01,0.17061760425567626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,2,power_law_1.01,2.395187187194824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,2,power_law_1.01,0.21448960304260253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,2,power_law_1.01,0.2421504020690918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,2,power_law_1.01,4.7575935363769535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,2,power_law_1.01,0.3344320058822632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,2,power_law_1.01,0.42632322311401366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,2,power_law_1.01,0.5669951915740967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,2,power_law_1.01,0.7286975860595704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,2,power_law_1.01,1.040601634979248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,2,power_law_1.01,1.3413439750671388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,2,power_law_1.01,1.7426368713378906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,2,power_law_1.01,2.53045768737793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,2,power_law_1.01,5.028339385986328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,2,power_law_1.01,0.05619199872016907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,2,power_law_1.01,0.051648002862930295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,2,power_law_1.01,0.06699519753456115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,2,power_law_1.01,0.07822719812393189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,2,power_law_1.01,0.08636800050735474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,2,power_law_1.01,0.12297600507736206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,2,power_law_1.01,0.12156800031661988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,2,power_law_1.01,0.12300159931182861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,2,power_law_1.01,0.12422399520874024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,2,power_law_1.01,0.12670079469680787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,2,power_law_1.01,0.12883199453353883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,2,power_law_1.01,0.12783360481262207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,2,power_law_1.01,0.13055360317230225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,2,power_law_1.01,0.13425920009613038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,2,power_law_1.01,0.14288640022277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,2,power_law_1.01,0.1468160033226013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,2,power_law_1.01,0.1538879990577698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,2,power_law_1.01,0.17902719974517822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.01,0.19158400297164918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.01,0.23199999332427979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.01,0.26820480823516846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.01,0.3639039993286133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.01,0.4328767776489258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.01,0.5897664070129395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.01,0.7991231918334961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.01,1.1023360252380372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.01,1.4882752418518066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,2,power_law_1.01,1.7643392562866211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,2,power_law_1.01,2.80185604095459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,2,power_law_1.01,5.702150344848633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,2,power_law_1.01,0.018822400271892546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,2,power_law_1.01,0.020102399587631225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,2,power_law_1.01,0.02338559925556183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,2,power_law_1.01,0.027692800760269164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,2,power_law_1.01,0.03154560029506683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,2,power_law_1.01,0.041433599591255185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,2,power_law_1.01,0.043724799156188966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,2,power_law_1.01,0.04741120040416717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,2,power_law_1.01,0.04747520089149475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,2,power_law_1.01,0.049465599656105044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,2,power_law_1.01,0.05061759948730469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,2,power_law_1.01,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,2,power_law_1.01,0.05434880256652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,2,power_law_1.01,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,2,power_law_1.01,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,2,power_law_1.01,0.05899519920349121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,2,power_law_1.01,0.06034560203552246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,2,power_law_1.01,0.10835839509963989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.01,0.11640959978103638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.01,0.11570559740066529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,2,power_law_1.2,0.05334399938583374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.01,0.12082560062408447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,2,power_law_1.2,0.05132160186767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.01,0.1428607940673828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,2,power_law_1.2,0.061286401748657224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.01,0.16365439891815187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,2,power_law_1.2,0.08009600043296813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.01,0.2149888038635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,2,power_law_1.2,0.08698880076408386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.01,0.2588927984237671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,2,power_law_1.2,0.13598719835281373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.01,0.35333120822906494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,2,power_law_1.2,0.14314240217208862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.01,0.4814335823059082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,2,power_law_1.2,0.14766720533370972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,2,power_law_1.01,0.5698495864868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,2,power_law_1.2,0.14832639694213867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,2,power_law_1.2,0.04568960070610047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,2,power_law_1.2,0.14883840084075928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,2,power_law_1.01,0.8232064247131348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,2,power_law_1.2,0.046137601137161255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,2,power_law_1.2,0.14547840356826783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,2,power_law_1.2,0.052223998308181765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,2,power_law_1.01,1.6203327178955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,2,power_law_1.2,0.1490623950958252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,2,power_law_1.2,0.05960320234298706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,2,power_law_1.2,0.1543552041053772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,2,power_law_1.2,0.06706560254096985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,2,power_law_1.2,0.1555008053779602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,2,power_law_1.2,0.1014847993850708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,2,power_law_1.2,0.16814080476760865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,2,power_law_1.2,0.10640000104904175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,2,power_law_1.2,0.16668800115585328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,2,power_law_1.2,0.10476800203323364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,2,power_law_1.2,0.17160320281982422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,2,power_law_1.2,0.10888320207595825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,2,power_law_1.2,0.10739840269088745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,2,power_law_1.2,0.1889855980873108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,2,power_law_1.2,0.11048959493637085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,2,power_law_1.2,0.20885119438171387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,2,power_law_1.2,0.11267199516296386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,2,power_law_1.2,0.25525119304656985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,2,power_law_1.2,0.11413120031356812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,2,power_law_1.2,0.2743040084838867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,2,power_law_1.2,0.11898239850997924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,2,power_law_1.2,0.3388351917266846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,2,power_law_1.2,0.12716799974441528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,2,power_law_1.2,0.439577579498291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,2,power_law_1.2,0.12849279642105102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,2,power_law_1.2,0.5918015956878662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,2,power_law_1.2,0.13778560161590575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,2,power_law_1.2,0.7901311874389648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,2,power_law_1.2,0.15941760540008545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,2,power_law_1.2,0.9949631690979004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,2,power_law_1.2,0.18087680339813234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,2,power_law_1.2,1.2673855781555177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,2,power_law_1.2,0.2166208028793335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,2,power_law_1.2,1.586911964416504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,2,power_law_1.2,0.25615360736846926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,2,power_law_1.2,0.3405888080596924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,2,power_law_1.2,2.6485055923461913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,2,power_law_1.2,0.41034879684448244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,2,power_law_1.2,5.470182418823242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,2,power_law_1.2,0.5874303817749024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,2,power_law_1.2,0.7584896087646484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,2,power_law_1.2,1.087564754486084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,2,power_law_1.2,1.4387328147888183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,2,power_law_1.2,1.6847999572753907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,2,power_law_1.2,2.935775947570801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,2,power_law_1.2,5.745862579345703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,2,power_law_1.2,0.05704960227012634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,2,power_law_1.2,0.05308799743652344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,2,power_law_1.2,0.06087679862976074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,2,power_law_1.2,0.07656959891319275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,2,power_law_1.2,0.08759040236473084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,2,power_law_1.2,0.11858559846878051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,2,power_law_1.2,0.128601598739624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,2,power_law_1.2,0.12584960460662842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,2,power_law_1.2,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,2,power_law_1.2,0.13112959861755372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,2,power_law_1.2,0.130457603931427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,2,power_law_1.2,0.13216639757156373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,2,power_law_1.2,0.13173760175704957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,2,power_law_1.2,0.13846399784088134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,2,power_law_1.2,0.14763519763946534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,2,power_law_1.2,0.1504320025444031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,2,power_law_1.2,0.15729279518127443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,2,power_law_1.2,0.17537920475006102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.2,0.20105600357055664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.2,0.23895680904388428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.2,0.2920896053314209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.2,0.3779711961746216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.2,0.4502528190612793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.2,0.6915071964263916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.2,0.813657569885254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.2,1.2172224044799804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.2,1.5304320335388184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,2,power_law_1.2,1.8322431564331054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,2,power_law_1.2,3.0199296951293944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,2,power_law_1.2,6.466169738769532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,2,power_law_1.2,0.019046400487422944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,2,power_law_1.2,0.020185600221157073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,2,power_law_1.2,0.022771200537681578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,2,power_law_1.2,0.027500799298286437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,2,power_law_1.2,0.030963200330734252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,2,power_law_1.2,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,2,power_law_1.2,0.04421760141849518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,2,power_law_1.2,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,2,power_law_1.2,0.04901759922504425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,2,power_law_1.2,0.050297600030899045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,2,power_law_1.2,0.05133439898490906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,2,power_law_1.2,0.05297279953956604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,2,power_law_1.2,0.0541055977344513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,2,power_law_1.2,0.05539199709892273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,2,power_law_1.2,0.0627839982509613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,2,power_law_1.2,0.06033920049667359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,1,balanced,0.05046399931112925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,2,power_law_1.2,0.06382719874382019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,1,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,2,power_law_1.2,0.11098239421844483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,1,balanced,0.07830399771531422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.2,0.11945600509643554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,1,balanced,0.11573867003122966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,1,balanced,0.17920533816019693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.2,0.12102400064468384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,1,balanced,0.17294400930404663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.2,0.12761600017547609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,1,balanced,0.16475733121236166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.2,0.14261759519577027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,1,balanced,0.1632426679134369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,1,balanced,0.1660266617933909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.2,0.16644480228424072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,1,balanced,0.1671839952468872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.2,0.22362880706787108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,1,balanced,0.16765334208806357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.2,0.2627583980560303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,1,balanced,0.1728853384653727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.2,0.3712831974029541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,1,balanced,0.17189866304397583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.2,0.45285758972167967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,1,balanced,0.17575999101003012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,2,power_law_1.2,0.5645311832427978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,1,balanced,0.18456000089645386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,1,balanced,0.048298666874567665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,1,balanced,0.05486399928728739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,1,balanced,0.18521066506703696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,2,power_law_1.2,0.8815808296203613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,1,balanced,0.0654666672150294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,1,balanced,0.19641600052515665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,1,balanced,0.09686932961146037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,2,power_law_1.2,1.7603647232055664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,1,balanced,0.13899200161298117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,1,balanced,0.21889066696166992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,1,balanced,0.13526933391888937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,1,balanced,0.13159466783205667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,1,balanced,0.23938665787378946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,1,balanced,0.13723199566205344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,1,balanced,0.13668266932169595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,1,balanced,0.2788693308830261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,1,balanced,0.1360053320725759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,1,balanced,0.13934933145840964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,1,balanced,0.3142186601956685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,1,balanced,0.1415786643822988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,1,balanced,0.14601600170135498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,1,balanced,0.14967466394106546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,1,balanced,0.4554400046666463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,1,balanced,0.15609600146611533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,1,balanced,0.1623840034008026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,1,balanced,0.17308266957600912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,1,balanced,0.5326240062713623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,1,balanced,0.19843200842539468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,1,balanced,0.2185386617978414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,1,balanced,0.2696053385734558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,1,balanced,0.7309066454569498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,1,balanced,0.3124799927075704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,1,balanced,0.46222933133443195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,1,balanced,0.9412639935811361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,1,balanced,0.5575733184814453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,1,balanced,1.3412106831868489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,1,balanced,0.795087973276774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,1,balanced,1.038256009419759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,1,balanced,1.7190613746643066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,1,power_law_1.01,0.04647679924964905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,1,power_law_1.01,0.05360640287399292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,1,power_law_1.01,0.06658560037612915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,1,power_law_1.01,0.0846015989780426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,1,power_law_1.01,0.10397440195083618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,1,balanced,1.513210614522298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,1,power_law_1.01,0.14767359495162963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,1,power_law_1.01,0.1539072036743164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,1,power_law_1.01,0.15436160564422607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,1,power_law_1.01,0.1564095973968506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,1,balanced,2.120453357696533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,1,power_law_1.01,0.15815039873123168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,1,power_law_1.01,0.1624511957168579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,1,power_law_1.01,0.16536320447921754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,1,power_law_1.01,0.1693567991256714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,1,power_law_1.01,0.1727552056312561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,1,power_law_1.01,0.17905919551849364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,1,power_law_1.01,0.18469760417938233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,1,power_law_1.01,0.19432320594787597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,1,power_law_1.01,0.22376320362091065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,1,balanced,1.9780319531758626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,1,power_law_1.01,0.2556864023208618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,1,power_law_1.01,0.29901440143585206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,1,power_law_1.01,0.34771840572357177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,1,power_law_1.01,0.4510079860687256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,1,power_law_1.01,0.547276782989502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,1,power_law_1.01,0.7559296131134033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,1,power_law_1.01,0.9368127822875977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,1,power_law_1.01,1.330630397796631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,1,power_law_1.01,1.7216127395629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,1,balanced,3.2962398529052734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,1,power_law_1.01,2.1302528381347656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,1,power_law_1.01,3.307865524291992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,1,power_law_1.01,6.489459228515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,1,balanced,2.435370604197184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,1,power_law_1.01,0.043724799156188966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,1,power_law_1.01,0.049491199851036075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,1,power_law_1.01,0.05957760214805603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,1,power_law_1.01,0.07265920042991639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,1,power_law_1.01,0.09069439768791199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,1,power_law_1.01,0.11863679885864258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,1,power_law_1.01,0.12357759475708008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,1,power_law_1.01,0.12805119752883912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,1,power_law_1.01,0.12899839878082275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,1,power_law_1.01,0.13165440559387206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,1,power_law_1.01,0.13506560325622557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,1,power_law_1.01,0.13836159706115722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,1,balanced,3.821194648742676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,1,power_law_1.01,0.14265600442886353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,1,power_law_1.01,0.1458624005317688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,1,power_law_1.01,0.15495680570602416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,1,power_law_1.01,0.16082559823989867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,1,power_law_1.01,0.17402880191802977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,1,power_law_1.01,0.20347518920898439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,1,balanced,6.509397506713867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,1,power_law_1.01,0.2321216106414795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,1,power_law_1.01,0.29219200611114504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,1,power_law_1.01,0.3474623918533325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,1,power_law_1.01,0.4678336143493652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,1,power_law_1.01,0.5809728145599365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,1,power_law_1.01,0.8227904319763184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,1,power_law_1.01,1.0455103874206544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,1,power_law_1.01,1.5116288185119628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,1,power_law_1.01,1.97391357421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,1,power_law_1.01,2.437875175476074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,1,power_law_1.01,3.8242752075195314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,1,power_law_1.01,7.520941162109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,1,balanced,7.5282026926676435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,1,power_law_1.2,0.04677119851112366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,1,power_law_1.2,0.05363839864730835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,1,power_law_1.2,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,1,power_law_1.2,0.07868800163269044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,1,power_law_1.2,0.09096320271492005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,1,power_law_1.2,0.14577920436859132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,1,power_law_1.2,0.1527616024017334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,1,power_law_1.2,0.1578879952430725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,1,power_law_1.2,0.15904639959335326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,1,power_law_1.2,0.15924479961395263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,1,power_law_1.2,0.1661631941795349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,1,power_law_1.2,0.1704576015472412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,1,power_law_1.2,0.17107839584350587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,1,power_law_1.2,0.17365119457244874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,1,power_law_1.2,0.18443520069122316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,1,power_law_1.2,0.1893183946609497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,1,power_law_1.2,0.2001728057861328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,1,power_law_1.2,0.23489279747009278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,1,power_law_1.2,0.25500800609588625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,1,power_law_1.2,0.30127360820770266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,1,power_law_1.2,0.34875519275665284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,1,power_law_1.2,0.450816011428833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,1,power_law_1.2,0.5476352214813233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,1,power_law_1.2,0.7487552165985107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,1,power_law_1.2,0.9367551803588867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,1,power_law_1.2,1.325817584991455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,1,power_law_1.2,1.7219520568847657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,1,power_law_1.2,2.126380729675293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,1,power_law_1.2,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,1,power_law_1.2,0.0506496012210846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,1,power_law_1.2,3.3036609649658204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,1,power_law_1.2,0.0581055998802185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,1,power_law_1.2,6.4801788330078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,1,power_law_1.2,0.06740480065345764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,1,power_law_1.2,0.08707200288772583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,1,power_law_1.2,0.12133760452270508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,1,power_law_1.2,0.12616959810256959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,1,power_law_1.2,0.12800639867782593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,1,power_law_1.2,0.13100800514221192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,1,power_law_1.2,0.1331007957458496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,1,power_law_1.2,0.13285759687423707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,1,power_law_1.2,0.14039039611816406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,8,balanced,0.08917333682378133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,1,power_law_1.2,0.14332799911499022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,8,balanced,0.10267733534177144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,1,power_law_1.2,0.14855680465698243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,8,balanced,0.14380266269048056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,1,power_law_1.2,0.15594880580902098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,8,balanced,0.22156800826390585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,1,power_law_1.2,0.16286720037460328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,8,balanced,0.35441064834594727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,8,balanced,0.4296799898147583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,1,power_law_1.2,0.1743551969528198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,8,balanced,0.4244906504948934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,1,power_law_1.2,0.21052799224853516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,8,balanced,0.43598934014638263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,1,power_law_1.2,0.23566720485687256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,8,balanced,0.45396268367767334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,1,power_law_1.2,0.2934272050857544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,8,balanced,0.41154666741689044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,1,power_law_1.2,0.3469376087188721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,8,balanced,0.41994667053222656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,1,power_law_1.2,0.4665664196014404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,8,balanced,0.4400586684544881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,1,power_law_1.2,0.577945613861084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,8,balanced,0.4461866617202759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,8,balanced,0.4119946559270223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,1,power_law_1.2,0.8178560256958007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,8,balanced,0.43565332889556885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,1,power_law_1.2,1.0423616409301757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,8,balanced,0.40305598576863605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,1,power_law_1.2,1.5080831527709961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,8,balanced,0.425439993540446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,1,power_law_1.2,1.9705663681030274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,1,power_law_1.2,2.4312000274658203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,8,balanced,0.4171626567840576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,1,power_law_1.2,3.818368148803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,8,balanced,0.41319998105367023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,1,power_law_1.2,7.514035034179687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,8,balanced,0.40719465414683026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,8,balanced,0.06668800115585327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,8,balanced,0.07133866846561432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,8,balanced,0.4142719904581706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,8,balanced,0.10020266969998677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,8,balanced,0.14380799730618796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,8,balanced,0.22127999862035116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,8,balanced,0.25574400027592975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,8,balanced,0.6934239864349365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,8,balanced,0.23993066946665445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,8,balanced,0.2407146692276001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,8,balanced,0.25563732782999676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,8,balanced,0.2366186579068502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,8,balanced,0.6252426703770956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,8,balanced,0.2413653333981832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,8,balanced,0.24154667059580484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,8,balanced,0.25809067487716675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,8,balanced,0.2595146695772807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,8,balanced,0.9241600036621094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,8,balanced,0.25989866256713867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,8,balanced,0.25889599323272705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,8,balanced,0.25996800263722736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,8,balanced,0.26243199904759723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,8,balanced,1.2264800071716309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,8,balanced,0.26241066058476764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,8,balanced,0.27244265874226886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,8,balanced,0.29126934210459393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,8,balanced,1.2997600237528484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,8,balanced,0.4670720100402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,8,balanced,0.4403626521428426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,8,balanced,1.9158080418904622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,8,balanced,0.6508959929148356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,8,balanced,0.8525973161061605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,8,balanced,1.9285227457682292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,8,balanced,0.9921706517537435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,8,balanced,0.06433066725730896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,8,balanced,0.0755573312441508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,8,balanced,0.09499200185139973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,8,balanced,0.1413226624329885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,8,balanced,0.2416213353474935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,8,balanced,0.24099200963974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,8,balanced,0.2214453419049581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,8,balanced,0.23465599616368613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,8,balanced,0.2220159967740377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,8,balanced,0.2213546633720398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,8,balanced,0.23347200949986777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,8,balanced,0.25042132536570233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,8,balanced,0.2368160088857015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,8,balanced,0.24895467360814413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,8,balanced,0.2705120046933492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,8,balanced,1.4174453417460124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,8,balanced,0.24356265862782797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,8,balanced,0.23831466833750406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,8,balanced,3.1166346867879233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,8,balanced,0.2533973256746928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,8,balanced,0.23811733722686768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,8,balanced,0.24697067340215048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,8,balanced,0.2457546591758728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,8,balanced,0.4235680103302002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,8,balanced,0.4105653365453084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,8,balanced,1.5088799794514973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,8,balanced,0.5899093151092529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,8,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,8,balanced,0.045509333411852516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,8,balanced,0.05351999898751577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,8,balanced,0.06331199904282887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,8,balanced,0.08763733506202698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,8,balanced,0.7673226992289225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,8,balanced,0.10259200135866801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,8,balanced,0.12198932965596516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,8,balanced,0.11939199765523274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,8,balanced,0.12431466579437256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,8,balanced,0.1729066570599874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,8,balanced,0.16457600394884744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,8,balanced,0.1736533244450887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,8,balanced,0.21397866805394491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,8,balanced,0.2209119995435079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,8,balanced,0.8652053674062093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,8,balanced,0.24769065777460733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,8,balanced,0.23954667647679648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,8,balanced,0.22430932521820068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,8,balanced,0.24332799514134726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,8,balanced,0.26396799087524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,8,balanced,2.394714673360189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,8,balanced,6.135749181111653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,8,balanced,0.29917333523432416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,8,balanced,0.34080533186594647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,8,balanced,1.2499840259552002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,8,balanced,0.4005813201268514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,8,balanced,0.45934398969014484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,8,balanced,0.7187306880950928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,8,balanced,1.3131626447041829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,8,balanced,0.8597706953684489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,8,balanced,1.2632160186767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,8,balanced,2.0755839347839355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,8,balanced,1.6509706179300945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,8,balanced,4.77562141418457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,8,balanced,1.9045012791951497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,8,balanced,3.1542933781941733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,8,balanced,4.131221453348796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,8,power_law_1.01,0.21791999340057372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,8,power_law_1.01,0.23056640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,8,power_law_1.01,0.13918720483779906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,8,power_law_1.01,0.20139520168304442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,8,power_law_1.01,0.2610368013381958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,8,power_law_1.01,0.3415872097015381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,8,power_law_1.01,0.38613760471343994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,8,balanced,6.023663838704427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,8,power_law_1.01,0.3607039928436279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,8,power_law_1.01,0.3939199924468994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,8,power_law_1.01,0.3824448108673096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,8,power_law_1.01,0.3870399951934814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,8,power_law_1.01,0.38007678985595705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,8,power_law_1.01,0.41310720443725585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,8,power_law_1.01,0.4135104179382324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,8,power_law_1.01,0.42731518745422364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,8,power_law_1.01,0.39370880126953123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,8,power_law_1.01,0.43217921257019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,8,power_law_1.01,0.4674560070037842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,8,power_law_1.01,0.5000639915466308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,8,power_law_1.01,0.5847296237945556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,8,power_law_1.01,0.6399360179901123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,8,power_law_1.01,0.7928383827209473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,8,power_law_1.01,0.9493375778198242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,8,power_law_1.01,1.309823989868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,8,power_law_1.01,1.5392383575439452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,8,power_law_1.01,2.1188095092773436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,8,power_law_1.01,2.7549503326416014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,8,power_law_1.01,3.5542270660400392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,8,power_law_1.01,0.13415679931640626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,8,power_law_1.01,4.81014404296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,8,power_law_1.01,0.17221759557723998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,8,power_law_1.01,0.1076416015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,8,power_law_1.01,10.637728118896485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,8,power_law_1.01,0.11992319822311401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,8,power_law_1.01,0.14382079839706421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,8,power_law_1.01,0.21389439105987548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,8,power_law_1.01,0.22458879947662352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,8,power_law_1.01,0.22556159496307374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,8,power_law_1.01,0.24033920764923095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,8,power_law_1.01,0.22865920066833495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,8,power_law_1.01,0.2333888053894043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,8,power_law_1.01,0.24262399673461915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,8,power_law_1.01,0.24781439304351807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,8,power_law_1.01,0.2369920015335083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,8,power_law_1.01,0.24812800884246827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,8,power_law_1.01,0.25384318828582764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,8,power_law_1.01,0.2640575885772705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,8,power_law_1.01,0.2916863918304443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,8,power_law_1.01,0.3300607919692993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,8,power_law_1.01,0.3739392042160034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,8,power_law_1.01,0.4230207920074463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,8,power_law_1.01,0.5444992065429688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,8,power_law_1.01,0.6754816055297852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,8,power_law_1.01,0.8332287788391113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,8,power_law_1.01,0.9942591667175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,8,power_law_1.01,0.13933440446853637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,8,power_law_1.01,1.4170432090759277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,8,power_law_1.01,0.15245440006256103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,8,power_law_1.01,1.8585855484008789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,8,power_law_1.01,0.101254403591156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,8,power_law_1.01,2.273196792602539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,8,power_law_1.01,0.15186560153961182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,8,power_law_1.01,0.16253440380096434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,8,power_law_1.01,3.868115234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,8,power_law_1.01,0.19998719692230224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,8,power_law_1.01,6.840083312988281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,8,power_law_1.01,0.21692800521850586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,8,power_law_1.01,0.23505918979644774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,8,power_law_1.01,0.21812479496002196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,8,power_law_1.01,0.2358975887298584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,8,power_law_1.01,0.2356544017791748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,8,power_law_1.01,0.22853119373321534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,8,power_law_1.01,0.23651199340820311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,8,power_law_1.01,0.23682560920715331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,8,power_law_1.01,0.23856000900268554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,8,power_law_1.01,0.24095358848571777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,8,power_law_1.01,0.2585792064666748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,8,power_law_1.01,0.2865151882171631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.01,0.30433919429779055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.01,0.3554176092147827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.01,0.39219839572906495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.01,0.483187198638916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.01,0.5705215930938721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.01,0.7747712135314941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.01,0.9858816146850586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.01,1.419257640838623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.01,1.891801643371582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,8,power_law_1.01,2.018284797668457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,8,power_law_1.01,3.2529983520507812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,8,power_law_1.01,0.04991360008716583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,8,power_law_1.01,0.05086719989776611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,8,power_law_1.01,5.993344116210937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,8,power_law_1.01,0.045407998561859134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,8,power_law_1.01,0.05649920105934143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,8,power_law_1.01,0.05924479961395264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,8,power_law_1.01,0.08785920143127442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,8,power_law_1.01,0.09507840275764465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,8,power_law_1.01,0.09557759761810303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,8,power_law_1.01,0.09864959716796876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,8,power_law_1.01,0.1411520004272461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,8,power_law_1.01,0.14169600009918212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,8,power_law_1.01,0.12963199615478516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,8,power_law_1.01,0.183513605594635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,8,power_law_1.01,0.19164799451828002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,8,power_law_1.01,0.19443199634552003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,8,power_law_1.01,0.2139904022216797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,8,power_law_1.01,0.19883519411087036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,8,power_law_1.01,0.18789759874343873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.01,0.19815679788589477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.01,0.20752639770507814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.01,0.2266495943069458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.01,0.27934720516204836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.01,0.3052160024642944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.01,0.4523263931274414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.01,0.5546048164367676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.01,0.8081536293029785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.01,0.9862079620361328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,8,power_law_1.01,1.2418944358825683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,8,power_law_1.01,2.0265087127685546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,8,power_law_1.2,0.22268800735473632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,8,power_law_1.2,0.1612223982810974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,8,power_law_1.01,3.9304065704345703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,8,power_law_1.2,0.15387519598007202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,8,power_law_1.2,0.18280960321426393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,8,power_law_1.2,0.2363136053085327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,8,power_law_1.2,0.35466880798339845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,8,power_law_1.2,0.384115195274353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,8,power_law_1.2,0.381113600730896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,8,power_law_1.2,0.3765120029449463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,8,power_law_1.2,0.39388160705566405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,8,power_law_1.2,0.3917952060699463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,8,power_law_1.2,0.4203904151916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,8,power_law_1.2,0.4145664215087891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,8,power_law_1.2,0.41919360160827634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,8,power_law_1.2,0.42495360374450686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,8,power_law_1.2,0.4203648090362549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,8,power_law_1.2,0.456006383895874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,8,power_law_1.2,0.48545279502868655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,8,power_law_1.2,0.5280384063720703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,8,power_law_1.2,0.6246655941009521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,8,power_law_1.2,0.7018943786621094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,8,power_law_1.2,0.9239359855651855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,8,power_law_1.2,1.0936575889587403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,8,power_law_1.2,1.4622271537780762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,8,power_law_1.2,0.13654400110244752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,8,power_law_1.2,1.7159168243408203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,8,power_law_1.2,0.09080960154533387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,8,power_law_1.2,2.380057525634766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,8,power_law_1.2,0.0892799973487854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,8,power_law_1.2,0.11544320583343506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,8,power_law_1.2,3.0186431884765623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,8,power_law_1.2,0.14962559938430786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,8,power_law_1.2,4.048204803466797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,8,power_law_1.2,0.21247999668121337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,8,power_law_1.2,0.2136064052581787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,8,power_law_1.2,6.838784027099609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,8,power_law_1.2,0.21424639225006104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,8,power_law_1.2,0.2193727970123291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,8,power_law_1.2,13.923866271972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,8,power_law_1.2,0.22306559085845948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,8,power_law_1.2,0.23447039127349853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,8,power_law_1.2,0.240012788772583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,8,power_law_1.2,0.2464384078979492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,8,power_law_1.2,0.24785919189453126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,8,power_law_1.2,0.2471168041229248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,8,power_law_1.2,0.25998079776763916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,8,power_law_1.2,0.2691648006439209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,8,power_law_1.2,0.30519680976867675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,8,power_law_1.2,0.3362047910690308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,8,power_law_1.2,0.3855936050415039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,8,power_law_1.2,0.4719871997833252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,8,power_law_1.2,0.5624447822570801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,8,power_law_1.2,0.6795263767242432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,8,power_law_1.2,1.0167936325073241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,8,power_law_1.2,1.1923839569091796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,8,power_law_1.2,1.6879167556762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,8,power_law_1.2,2.264217567443848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,8,power_law_1.2,2.9160703659057616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,8,power_law_1.2,3.891936111450195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,8,power_law_1.2,0.1411903977394104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,8,power_law_1.2,0.08575360178947448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,8,power_law_1.2,7.855506896972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,8,power_law_1.2,0.10455039739608765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,8,power_law_1.2,0.1224511981010437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,8,power_law_1.2,0.14882559776306153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,8,power_law_1.2,0.2244863986968994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,8,power_law_1.2,0.20974719524383545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,8,power_law_1.2,0.215283203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,8,power_law_1.2,0.2208575963973999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,8,power_law_1.2,0.2240895986557007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,8,power_law_1.2,0.23573760986328124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,8,power_law_1.2,0.24281599521636962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,8,power_law_1.2,0.24732799530029298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,8,power_law_1.2,0.23196160793304443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,8,power_law_1.2,0.24812800884246827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,8,power_law_1.2,0.24996480941772461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,8,power_law_1.2,0.28243839740753174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,8,power_law_1.2,0.29839999675750734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.2,0.32103040218353274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.2,0.3836927890777588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.2,0.4279359817504883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.2,0.5455167770385743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.2,0.6731967926025391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,8,power_law_1.2,0.0500544011592865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.2,0.9245311737060546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,8,power_law_1.2,0.030963200330734252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.2,1.0996992111206054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,8,power_law_1.2,0.045638400316238406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.2,1.4012288093566894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,8,power_law_1.2,0.05607680082321167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.2,2.071206474304199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,8,power_law_1.2,0.062483197450637816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,8,power_law_1.2,0.08616960048675537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,8,power_law_1.2,2.7196544647216796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,8,power_law_1.2,0.09871360063552856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,8,power_law_1.2,3.69249267578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,8,power_law_1.2,0.10285439491271972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,8,power_law_1.2,0.10181119441986083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,8,power_law_1.2,7.980467224121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,8,power_law_1.2,0.13814400434494017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,8,power_law_1.2,0.14007680416107177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,8,power_law_1.2,0.1224128007888794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,8,power_law_1.2,0.18788479566574096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,8,power_law_1.2,0.18791680335998534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,8,power_law_1.2,0.20967040061950684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,8,power_law_1.2,0.19572479724884034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,8,power_law_1.2,0.19775359630584716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,8,power_law_1.2,0.19715839624404907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.2,0.18401919603347777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.2,0.21750400066375733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.2,0.2438271999359131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.2,0.2881407976150513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.2,0.3729664087295532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.2,0.5001791954040528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.2,0.6259583950042724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.2,0.9181376457214355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.2,1.1640512466430664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,8,power_law_1.2,1.463270378112793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,8,power_law_1.2,2.355699157714844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,8,power_law_1.2,5.021843338012696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,4,balanced,0.0761653333902359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,4,balanced,0.09411199887593587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,4,balanced,0.13798399766286215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,4,balanced,0.20569066206614176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,4,balanced,0.3434293270111084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,4,balanced,0.41861867904663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,4,balanced,0.4010293483734131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,4,balanced,0.4014293352762858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,4,balanced,0.39762131373087567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,4,balanced,0.4002293348312378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,4,balanced,0.3858559926350911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,4,balanced,0.3810186783472697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,4,balanced,0.4188106854756673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,4,balanced,0.39903465906778973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,4,balanced,0.4014773368835449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,4,balanced,0.41688533624013263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,4,balanced,0.40199466546376544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,4,balanced,0.399455984433492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,4,balanced,0.4098293383916219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,4,balanced,0.4136213461558024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,4,balanced,0.43304534753163654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,4,balanced,0.7008372942606608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,4,balanced,0.060405333836873375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,4,balanced,0.06568000217278798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,4,balanced,0.6413493156433105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,4,balanced,0.09282666444778442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,4,balanced,0.13035200039545694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,4,balanced,0.20886399348576865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,4,balanced,0.22788800795873007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,4,balanced,0.22636266549428305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,4,balanced,0.22568533817927042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,4,balanced,0.9535466829935709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,4,balanced,1.2549866835276287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,4,balanced,1.4127999941507976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,4,balanced,0.0645546664794286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,4,balanced,2.0158185958862305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,4,balanced,0.07274133463700612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,4,balanced,0.0921013355255127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,4,balanced,0.14987732966740927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,4,balanced,0.22167466084162393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,4,balanced,0.2481173276901245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,4,balanced,0.23402132590611777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,4,balanced,0.23493866125742593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,4,balanced,0.22029866774876913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,4,balanced,0.2316533327102661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,4,balanced,0.22452267011006674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,4,balanced,0.22145066658655801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,4,balanced,0.2328640023867289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,4,balanced,0.2365600069363912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,4,balanced,0.23648534218470255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,4,balanced,0.23588265975316366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,4,balanced,2.102778593699137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,4,balanced,0.24664533138275146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,4,balanced,0.24465066194534302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,4,balanced,0.24345600605010986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,4,balanced,0.2598666747411092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,4,balanced,0.027104000250498455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,4,balanced,0.03230399886767069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,4,balanced,0.0466186652580897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,4,balanced,0.2746453285217285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,4,balanced,0.05942933261394501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,4,balanced,0.08555733164151509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,4,balanced,0.1042133371035258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,4,balanced,0.45054932435353595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,4,balanced,0.10379200180371602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,4,balanced,0.1081119974454244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,4,balanced,0.11198400457700093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,4,balanced,0.12043199936548869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,4,balanced,0.45424532890319824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,4,balanced,0.1272533337275187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,4,balanced,0.1284160017967224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,4,balanced,0.1688906749089559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,4,balanced,0.17427200078964233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,4,balanced,0.17149867614110312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,4,balanced,0.6491466760635376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,4,balanced,3.4171199798583984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,4,balanced,0.22312533855438232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,4,balanced,0.23737066984176636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,4,balanced,0.23883734146753946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,4,balanced,0.2552799979845683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,4,balanced,0.8516106605529785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,4,balanced,0.27901333570480347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,4,balanced,0.2916693290074666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,4,balanced,0.32710933685302734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,4,balanced,1.0332372983296711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,4,balanced,0.34779198964436847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,4,balanced,0.5506933530171713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,4,balanced,1.4308746655782063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,4,balanced,0.6250613530476888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,4,balanced,0.9122827053070068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,4,balanced,1.5976053873697917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,4,balanced,6.821301142374675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,4,balanced,1.1871679623921711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,4,balanced,0.23007466395696005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,4,balanced,0.2286240061124166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,4,balanced,0.23138666152954102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,4,balanced,0.23812266190846762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,4,balanced,0.2453333338101705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,4,balanced,0.23988266785939535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,4,balanced,2.5199039777119956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,4,balanced,0.24157333374023438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,4,balanced,1.3208746910095215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,4,balanced,0.25253866116205853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,4,balanced,0.24335465828577676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,4,balanced,0.266810675462087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,4,balanced,0.2643839915593465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,4,balanced,0.2908373276392619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,4,balanced,0.3165973424911499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,4,balanced,0.4700746536254883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,4,balanced,0.48976532618204754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,4,balanced,2.1977972984313965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,4,balanced,0.7081066767374674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,4,balanced,0.9172159830729166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,4,balanced,4.9686934153238935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,4,balanced,1.1403413613637288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,4,balanced,1.575642744700114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,4,balanced,4.1545759836832685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,4,balanced,1.769360065460205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,4,power_law_1.01,0.1469823956489563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,4,power_law_1.01,0.16508159637451172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,4,power_law_1.01,0.12723840475082399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,4,power_law_1.01,0.16747519969940186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,4,power_law_1.01,0.21916160583496094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,4,power_law_1.01,0.33590400218963623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,4,power_law_1.01,0.34394240379333496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,4,power_law_1.01,0.3582144021987915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,4,power_law_1.01,0.36124799251556394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,4,power_law_1.01,0.36827518939971926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,4,power_law_1.01,0.37009921073913576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,4,power_law_1.01,0.3729471921920776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,4,power_law_1.01,0.3781440019607544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,4,power_law_1.01,0.36932480335235596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,4,power_law_1.01,0.4016128063201904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,4,power_law_1.01,0.3787519931793213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,4,power_law_1.01,0.4165247917175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,4,power_law_1.01,0.4351359844207764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,4,power_law_1.01,0.487775993347168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,4,balanced,2.811744054158529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,4,power_law_1.01,0.5557759761810303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,4,power_law_1.01,0.5982016086578369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,4,power_law_1.01,0.7272448062896728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,4,power_law_1.01,0.9064319610595704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,4,power_law_1.01,1.1442367553710937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,4,power_law_1.01,1.4901951789855956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,4,power_law_1.01,2.0632896423339844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,4,power_law_1.01,2.3741504669189455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,4,power_law_1.01,2.656595230102539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,4,power_law_1.01,4.298163223266601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,4,power_law_1.01,8.310899353027343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,4,power_law_1.01,0.09639679789543151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,4,power_law_1.01,0.11610239744186401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,4,power_law_1.01,0.08891519904136658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,4,power_law_1.01,0.11372159719467163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,4,power_law_1.01,0.13693439960479736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,4,power_law_1.01,0.18944640159606935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,4,power_law_1.01,0.20171520709991456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,4,power_law_1.01,0.2092672109603882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,4,power_law_1.01,0.20773758888244628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,4,power_law_1.01,0.21182079315185548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,4,power_law_1.01,0.2214656114578247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,4,power_law_1.01,0.22088320255279542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,4,power_law_1.01,0.22791678905487062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,4,power_law_1.01,0.22222719192504883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,4,power_law_1.01,0.24051198959350586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,4,power_law_1.01,0.24078719615936278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,4,balanced,5.56923230489095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,4,power_law_1.01,0.2522495985031128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,4,power_law_1.01,0.2884416103363037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,4,power_law_1.01,0.3106559991836548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,4,power_law_1.01,0.36400001049041747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,4,power_law_1.01,0.413043212890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,4,power_law_1.01,0.5198336124420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,4,power_law_1.01,0.10726399421691894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,4,power_law_1.01,0.6173952102661133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,4,power_law_1.01,0.10192639827728271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,4,power_law_1.01,0.8346048355102539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,4,power_law_1.01,0.09311360120773315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,4,power_law_1.01,0.9797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,4,power_law_1.01,0.1218176007270813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,4,power_law_1.01,0.15639679431915282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,4,power_law_1.01,1.428268814086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,4,power_law_1.01,0.21719679832458497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,4,power_law_1.01,1.8160640716552734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,4,power_law_1.01,0.21076478958129882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,4,power_law_1.01,2.331884765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,4,power_law_1.01,0.22145919799804686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,4,power_law_1.01,3.313254547119141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,4,power_law_1.01,0.22787840366363527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,4,power_law_1.01,0.2292992115020752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,4,power_law_1.01,6.878880310058594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,4,power_law_1.01,0.23148798942565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,4,power_law_1.01,0.23568639755249024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,4,power_law_1.01,0.23519361019134521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,4,power_law_1.01,0.24065279960632324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,4,power_law_1.01,0.2311487913131714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,4,power_law_1.01,0.24196479320526124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,4,power_law_1.01,0.2532671928405762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,4,power_law_1.01,0.27915520668029786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.01,0.3054975986480713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.01,0.35470080375671387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.01,0.398086404800415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.01,0.4958911895751953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.01,0.5777152061462403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.01,0.8112640380859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.01,0.9543999671936035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.01,1.2607551574707032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.01,1.8362560272216797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,4,power_law_1.01,2.0249664306640627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,4,power_law_1.01,3.18408317565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,4,power_law_1.01,5.817023849487304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,4,power_law_1.01,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,4,power_law_1.01,0.04357759952545166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,4,power_law_1.01,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,4,power_law_1.01,0.04904319941997528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,4,power_law_1.01,0.06227840185165405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,4,power_law_1.01,0.080595201253891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,4,power_law_1.01,0.08697599768638611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,4,power_law_1.01,0.08940160274505615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,4,power_law_1.01,0.09333119988441467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,4,power_law_1.01,0.09672319889068604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,4,power_law_1.01,0.10047359466552734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,4,power_law_1.01,0.1017151951789856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,4,power_law_1.01,0.12137600183486938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,4,power_law_1.01,0.11528960466384888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,4,power_law_1.01,0.11605119705200195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,4,power_law_1.01,0.19328000545501708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,4,power_law_1.01,0.19718400239944459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,4,power_law_1.01,0.18065919876098632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.01,0.18673919439315795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.01,0.18995200395584105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.01,0.2082495927810669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.01,0.24941439628601075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.01,0.2993407964706421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.01,0.3644864082336426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,4,power_law_1.2,0.1386240005493164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.01,0.44759039878845214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,4,power_law_1.2,0.09364479780197144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.01,0.6325247764587403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,4,power_law_1.2,0.1294592022895813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,4,power_law_1.2,0.16597119569778443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.01,0.7822976112365723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,4,power_law_1.2,0.20490880012512208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,4,power_law_1.01,1.016921615600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,4,power_law_1.2,0.33727359771728516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,4,power_law_1.01,1.6487808227539062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,4,power_law_1.2,0.34461441040039065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,4,power_law_1.2,0.35420799255371094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,4,power_law_1.01,2.8614015579223633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,4,power_law_1.2,0.36037120819091795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,4,power_law_1.2,0.3683648109436035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,4,power_law_1.2,0.37344000339508054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,4,power_law_1.2,0.39322240352630616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,4,power_law_1.2,0.09595519900321961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,4,power_law_1.2,0.39118080139160155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,4,power_law_1.2,0.07679359912872315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,4,power_law_1.2,0.3894783973693848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,4,power_law_1.2,0.08116480112075805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,4,power_law_1.2,0.4142784118652344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,4,power_law_1.2,0.11282559633255004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,4,power_law_1.2,0.3985599994659424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,4,power_law_1.2,0.13559039831161498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,4,power_law_1.2,0.42062082290649416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,4,power_law_1.2,0.19617279767990112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,4,power_law_1.2,0.4684095859527588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,4,power_law_1.2,0.20737919807434083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,4,power_law_1.2,0.48696317672729494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,4,power_law_1.2,0.21420800685882568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,4,power_law_1.2,0.5641280174255371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,4,power_law_1.2,0.21509759426116942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,4,power_law_1.2,0.6353727817535401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,4,power_law_1.2,0.21607038974761963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,4,power_law_1.2,0.8204416275024414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,4,power_law_1.2,0.22884480953216552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,4,power_law_1.2,0.9108672142028809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,4,power_law_1.2,0.2250432014465332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,4,power_law_1.2,1.1932607650756837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,4,power_law_1.2,0.2332159996032715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,4,power_law_1.2,1.6603200912475586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,4,power_law_1.2,0.22957439422607423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,4,power_law_1.2,2.104582405090332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,4,power_law_1.2,0.2524672031402588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,4,power_law_1.2,2.7162496566772463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,4,power_law_1.2,0.2526655912399292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,4,power_law_1.2,0.2576832056045532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,4,power_law_1.2,3.0622783660888673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,4,power_law_1.2,0.2911231994628906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,4,power_law_1.2,4.906592178344726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,4,power_law_1.2,0.33418879508972166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,4,power_law_1.2,0.3672192096710205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,4,power_law_1.2,10.641849517822266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,4,power_law_1.2,0.09944959878921508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,4,power_law_1.2,0.4245759963989258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,4,power_law_1.2,0.09939200282096863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,4,power_law_1.2,0.5425280094146728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,4,power_law_1.2,0.09171839952468872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,4,power_law_1.2,0.6376192092895507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,4,power_law_1.2,0.1233407974243164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,4,power_law_1.2,0.9078399658203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,4,power_law_1.2,0.13994239568710326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,4,power_law_1.2,1.0855423927307128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,4,power_law_1.2,0.21777279376983644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,4,power_law_1.2,1.6667455673217773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,4,power_law_1.2,0.22771201133728028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,4,power_law_1.2,2.095795249938965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,4,power_law_1.2,0.2287424087524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,4,power_law_1.2,2.4631935119628907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,4,power_law_1.2,0.2212671995162964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,4,power_law_1.2,0.24021120071411134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,4,power_law_1.2,4.342681503295898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,4,power_law_1.2,0.23732481002807618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,4,power_law_1.2,8.064198303222657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,4,power_law_1.2,0.2383039951324463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,4,power_law_1.2,0.2363584041595459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,4,power_law_1.2,0.2453376054763794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,4,power_law_1.2,0.24700798988342285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,4,power_law_1.2,0.2466559886932373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,4,power_law_1.2,0.2624255895614624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,4,power_law_1.2,0.2863231897354126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.2,0.3275840044021606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.2,0.3638783931732178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.2,0.43902077674865725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.2,0.4905087947845459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.2,0.6193600177764893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.2,0.8533632278442382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.2,1.0261119842529296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.2,1.4833984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.2,1.9815807342529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,4,power_law_1.2,2.385696029663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,4,power_law_1.2,3.5897918701171876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,4,power_law_1.2,7.849273681640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,4,power_law_1.2,0.03182080090045929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,4,power_law_1.2,0.028998398780822755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,4,power_law_1.2,0.036422398686408994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,4,power_law_1.2,0.047705599665641786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,4,power_law_1.2,0.058323198556900026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,4,power_law_1.2,0.08433279991149903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,4,power_law_1.2,0.08482559919357299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,4,power_law_1.2,0.09122560024261475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,4,power_law_1.2,0.0934656023979187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,4,power_law_1.2,0.09624320268630981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,4,power_law_1.2,0.10024960041046142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,4,power_law_1.2,0.10118399858474732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,4,power_law_1.2,0.12569600343704224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,4,power_law_1.2,0.11553280353546143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,4,power_law_1.2,0.11568640470504761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,4,power_law_1.2,0.19486720561981202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,4,power_law_1.2,0.21144320964813232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,4,power_law_1.2,0.1903040051460266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.2,0.19108480215072632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.2,0.19461120367050172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.2,0.21393918991088867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.2,0.262009596824646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.2,0.30344319343566895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.2,0.4057024002075195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,2,balanced,0.07310399909814198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.2,0.49570560455322266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,2,balanced,0.095551997423172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,2,balanced,0.1453279952208201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.2,0.6745024204254151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,2,balanced,0.21803200244903564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.2,0.9232128143310547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,2,balanced,0.34644798437754315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,2,balanced,0.39363733927408856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,4,power_law_1.2,1.0822336196899414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,2,balanced,0.38467733065287274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,4,power_law_1.2,1.6396991729736328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,2,balanced,0.37652798493703205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,2,balanced,0.38126933574676514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,2,balanced,0.05875733494758606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,4,power_law_1.2,3.2356990814208983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,2,balanced,0.38359999656677246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,2,balanced,0.07317333420117696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,2,balanced,0.3796693483988444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,2,balanced,0.09658132990201314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,2,balanced,0.3739946683247884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,2,balanced,0.135861337184906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,2,balanced,0.3782240152359009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,2,balanced,0.20678400993347168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,2,balanced,0.2287999987602234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,2,balanced,0.3925173282623291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,2,balanced,0.22459199031194052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,2,balanced,0.3903733491897583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,2,balanced,0.2302186687787374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,2,balanced,0.408896009127299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,2,balanced,0.23203200101852417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,2,balanced,0.22878400484720865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,2,balanced,0.38996267318725586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,2,balanced,0.23139200607935587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,2,balanced,0.417738676071167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,2,balanced,0.2349546750386556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,2,balanced,0.23401600122451782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,2,balanced,0.42050667603810626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,2,balanced,0.24754667282104492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,2,balanced,0.24542933702468872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,2,balanced,0.45295464992523193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,2,balanced,0.24828267097473145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,2,balanced,0.25826666752497357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,2,balanced,0.48545066515604657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,2,balanced,0.28405332565307617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,2,balanced,0.296506663163503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,2,balanced,0.7177279790242513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,2,balanced,0.3407466808954875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,2,balanced,0.07285333176453908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,2,balanced,0.38036266962687176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,2,balanced,0.08601599931716919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,2,balanced,0.7289439837137858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,2,balanced,0.11539199948310852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,2,balanced,0.16237333416938782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,2,balanced,0.2418880065282186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,2,balanced,0.5424426794052124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,2,balanced,0.27215466896692914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,2,balanced,0.27275200684865314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,2,balanced,0.2585013310114543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,2,balanced,0.2592639923095703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,2,balanced,0.6056480010350546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,2,balanced,1.0627573331197102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,2,balanced,0.25878934065500897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,2,balanced,0.2548639973004659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,2,balanced,0.2704213261604309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,2,balanced,0.26633600393931073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,2,balanced,0.26075732707977295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,2,balanced,0.8702879746754965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,2,balanced,0.27404266595840454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,2,balanced,0.27287999788920086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,2,balanced,1.3785600662231445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,2,balanced,0.2835093339284261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,2,balanced,0.2969226638476054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,2,balanced,1.1226613521575928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,2,balanced,0.30642666419347125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,2,balanced,0.33952001730600995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,2,balanced,0.37167465686798096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,2,balanced,1.6677066485087078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,2,balanced,1.4788533846537273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,2,balanced,0.5894933144251505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,2,balanced,0.6431573232014974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,2,balanced,0.8996480305989584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,2,balanced,1.9739999771118164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,2,balanced,2.3093172709147134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,2,balanced,1.1795946756998699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,2,balanced,1.5276106198628743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,2,balanced,2.309216022491455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,2,balanced,2.608037312825521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,2,balanced,2.08681058883667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,2,balanced,3.699626604715983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,2,balanced,2.422111988067627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,2,balanced,0.025546667476495106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,2,balanced,0.030207999050617218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,2,balanced,0.04240000247955322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,2,balanced,0.05895466605822245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,2,balanced,0.09008000294367473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,2,balanced,0.1053653359413147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,2,balanced,0.11186133821805318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,2,balanced,0.1167733371257782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,2,balanced,0.12146666646003723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,2,balanced,4.197973251342773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,2,balanced,0.1132319966952006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,2,balanced,0.11701333522796631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,2,balanced,0.1246613363424937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,2,balanced,0.12943999965985617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,2,balanced,0.13140267133712769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,2,balanced,0.13397333025932312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,2,balanced,0.172650674978892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,2,balanced,0.17171732584635416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,2,balanced,0.2484053373336792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,2,balanced,0.2791680097579956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,2,balanced,0.2768213351567586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,2,balanced,3.8245280583699546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,2,balanced,0.28730666637420654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,2,balanced,0.3111306627591451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,2,balanced,0.3201333284378052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,2,balanced,7.335237503051758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,2,balanced,0.5276639858881632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,2,balanced,0.5817600091298422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,2,balanced,0.8321866989135742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,2,balanced,8.317370732625326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,2,balanced,7.542266845703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,2,balanced,1.0819466908772786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,2,balanced,1.1796639760335286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,2,balanced,1.9543840090433757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,2,power_law_1.01,0.09538559913635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,2,power_law_1.01,0.11691520214080811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,2,power_law_1.01,0.10611200332641602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,2,power_law_1.01,0.16057599782943727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,2,power_law_1.01,0.20234880447387696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,2,power_law_1.01,0.30351359844207765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,2,power_law_1.01,0.3182015895843506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,2,power_law_1.01,0.3282687902450562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,2,power_law_1.01,0.33286399841308595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,2,power_law_1.01,0.3548095941543579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,2,power_law_1.01,0.34973440170288084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,2,power_law_1.01,0.3562304019927979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,2,power_law_1.01,0.3607424020767212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,2,power_law_1.01,0.36142079830169677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,2,power_law_1.01,0.36858880519866943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,2,power_law_1.01,0.3806015968322754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,2,balanced,3.6833651860555015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,2,power_law_1.01,0.40200319290161135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,2,power_law_1.01,0.42766718864440917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,2,power_law_1.01,0.4609663963317871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,2,power_law_1.01,0.5249663829803467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,2,power_law_1.01,0.5822976112365723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,2,power_law_1.01,0.7450047969818115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,2,power_law_1.01,0.07271040081977845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,2,power_law_1.01,0.8778304100036621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,2,power_law_1.01,0.09637119770050048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,2,power_law_1.01,1.1344127655029297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,2,power_law_1.01,0.08256000280380249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,2,power_law_1.01,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,2,power_law_1.01,1.4009344100952148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,2,power_law_1.01,0.11210880279541016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,2,power_law_1.01,0.09728639721870422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,2,power_law_1.01,1.948953628540039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,2,power_law_1.01,0.09245439767837524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,2,power_law_1.01,0.13856639862060546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,2,power_law_1.01,0.13084160089492797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,2,power_law_1.01,0.1870144009590149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,2,power_law_1.01,2.530892753601074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,2,power_law_1.01,0.1683583974838257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,2,power_law_1.01,0.2004607915878296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,2,power_law_1.01,2.9361791610717773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,2,power_law_1.01,0.23272960186004638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,2,power_law_1.01,0.2096384048461914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,2,power_law_1.01,4.650713729858398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,2,power_law_1.01,0.2385727882385254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,2,power_law_1.01,0.21270399093627929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,2,power_law_1.01,0.23551359176635742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,2,power_law_1.01,0.2088383913040161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,2,power_law_1.01,9.105804443359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,2,power_law_1.01,0.24581120014190674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,2,power_law_1.01,0.21799681186676026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,2,power_law_1.01,0.24172799587249755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,2,power_law_1.01,0.2314687967300415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,2,power_law_1.01,0.2285759925842285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,2,power_law_1.01,0.24161279201507568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,2,power_law_1.01,0.23812479972839357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,2,power_law_1.01,0.24792320728302003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,2,power_law_1.01,0.2574143886566162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,2,power_law_1.01,0.2926208019256592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,2,power_law_1.01,0.3243583917617798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,2,power_law_1.01,0.3753727912902832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,2,power_law_1.01,0.43281922340393064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,2,power_law_1.01,0.5560704231262207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,2,power_law_1.01,0.6798399925231934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,2,power_law_1.01,0.916710376739502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,2,power_law_1.01,1.2049344062805176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,2,power_law_1.01,1.6175167083740234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,2,power_law_1.01,2.080659294128418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,2,power_law_1.01,2.5775232315063477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,2,power_law_1.01,4.155295944213867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,2,power_law_1.01,7.929100799560547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,2,power_law_1.01,0.025696000456809996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,2,power_law_1.01,0.031142398715019226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,2,power_law_1.01,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,2,power_law_1.01,0.04772480130195618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,2,power_law_1.01,0.05922560095787048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,2,power_law_1.01,0.08030719757080078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,2,power_law_1.01,0.08759040236473084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,2,power_law_1.01,0.09523839950561523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,2,power_law_1.2,0.09639679789543151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,2,power_law_1.01,0.09953920245170593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,2,power_law_1.2,0.08752639889717102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,2,power_law_1.01,0.09394559860229493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,2,power_law_1.2,0.11901439428329467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,2,power_law_1.01,0.10139520168304443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,2,power_law_1.2,0.15803519487380982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,2,power_law_1.01,0.10291839838027954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,2,power_law_1.2,0.1859007954597473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,2,power_law_1.01,0.10305279493331909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,2,power_law_1.2,0.3012095928192139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,2,power_law_1.01,0.10694400072097779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,2,power_law_1.2,0.32807040214538574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,2,power_law_1.01,0.11228799819946289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,2,power_law_1.2,0.33112320899963377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,2,power_law_1.01,0.11450239419937133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,2,power_law_1.01,0.12862720489501953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,2,power_law_1.2,0.3337088108062744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,2,power_law_1.01,0.2209088087081909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,2,power_law_1.2,0.35481600761413573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.01,0.2378175973892212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,2,power_law_1.2,0.36058878898620605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.01,0.20848000049591064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,2,power_law_1.2,0.36319360733032224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.01,0.22400639057159424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,2,power_law_1.2,0.366758394241333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,2,power_law_1.2,0.37370240688323975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.01,0.25080959796905516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,2,power_law_1.2,0.3964479923248291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.01,0.28952319622039796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,2,power_law_1.2,0.39792640209198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.01,0.3791167974472046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,2,power_law_1.2,0.40933761596679685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.01,0.4371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,2,power_law_1.2,0.45386881828308107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.01,0.6032192230224609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,2,power_law_1.2,0.47481598854064944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.01,0.8092800140380859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,2,power_law_1.2,0.5411456108093262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,2,power_law_1.01,0.9655872344970703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,2,power_law_1.2,0.6240255832672119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,2,power_law_1.01,0.2577791929244995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,2,power_law_1.01,1.478668785095215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,2,power_law_1.2,0.7564223766326904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,2,power_law_1.01,0.25638399124145506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,2,power_law_1.01,0.2666624069213867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,2,power_law_1.2,0.9038528442382813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,2,power_law_1.01,3.0070144653320314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,2,power_law_1.01,0.26287360191345216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,2,power_law_1.2,1.1514047622680663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,2,power_law_1.01,0.27409920692443845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,2,power_law_1.2,1.520633602142334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,2,power_law_1.01,0.2745856046676636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,2,power_law_1.2,2.0742656707763674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,2,power_law_1.01,0.2919552087783813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,2,power_law_1.2,0.07104640007019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,2,power_law_1.2,2.512339210510254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,2,power_law_1.01,0.319814395904541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,2,power_law_1.2,0.07093759775161743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.01,0.34028160572052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,2,power_law_1.2,3.200672149658203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,2,power_law_1.2,0.08650239706039428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.01,0.40542078018188477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,2,power_law_1.2,0.1109760046005249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,2,power_law_1.2,4.694035339355469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.01,0.4681856155395508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,2,power_law_1.2,0.12877440452575684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.01,0.6028736114501954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,2,power_law_1.2,0.1975551962852478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,2,power_law_1.2,10.229798126220704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,2,power_law_1.2,0.20387840270996094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.01,0.7507584095001221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,2,power_law_1.2,0.21137919425964355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.01,1.0281408309936524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,2,power_law_1.2,0.21251199245452881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.01,1.290835189819336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,2,power_law_1.2,0.2233151912689209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.01,1.7689216613769532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,2,power_law_1.2,0.23128321170806884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.01,2.2577728271484374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,2,power_law_1.2,0.23912320137023926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,2,power_law_1.01,2.7470783233642577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,2,power_law_1.2,0.2337280035018921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,2,power_law_1.2,0.2375999927520752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,2,power_law_1.01,4.341030502319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,2,power_law_1.2,0.2549184083938599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,2,power_law_1.2,0.25793919563293455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,2,power_law_1.01,8.37220458984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,2,power_law_1.2,0.26883840560913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,2,power_law_1.2,0.2961087942123413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,2,power_law_1.2,0.3367039918899536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,2,power_law_1.2,0.3996608018875122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,2,power_law_1.2,0.4527167797088623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,2,power_law_1.2,0.5669951915740967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,2,power_law_1.2,0.7170176029205322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,2,power_law_1.2,0.9382975578308106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,2,power_law_1.2,1.270521640777588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,2,power_law_1.2,1.6837696075439452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,2,power_law_1.2,2.2324287414550783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,2,power_law_1.2,2.7202367782592773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,2,power_law_1.2,4.06426887512207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,2,power_law_1.2,8.661190032958984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,2,power_law_1.2,0.08669440150260925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,2,power_law_1.2,0.08360959887504578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,2,power_law_1.2,0.09639679789543151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,2,power_law_1.2,0.12607359886169434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,2,power_law_1.2,0.14689279794692994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,2,power_law_1.2,0.2322240114212036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,2,power_law_1.2,0.2417151927947998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,2,power_law_1.2,0.24312961101531982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,2,power_law_1.2,0.25040640830993655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,2,power_law_1.2,0.25541760921478274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,2,power_law_1.2,0.25306239128112795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,2,power_law_1.2,0.2610687971115112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,2,power_law_1.2,0.2682687997817993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,2,power_law_1.2,0.022892799973487855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,2,power_law_1.2,0.27382400035858157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,2,power_law_1.2,0.025676798820495606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,2,power_law_1.2,0.2838144063949585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,2,power_law_1.2,0.03258880078792572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,2,power_law_1.2,0.29663360118865967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,2,power_law_1.2,0.04496000111103058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,2,power_law_1.2,0.29877119064331054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,2,power_law_1.2,0.054713600873947145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,2,power_law_1.2,0.3355776071548462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,2,power_law_1.2,0.08061439990997314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.2,0.35970559120178225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,2,power_law_1.2,0.09045119881629944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,2,power_law_1.2,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.2,0.4303936004638672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,1,balanced,0.08104533453782399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,2,power_law_1.2,0.10154240131378174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.2,0.48172798156738283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,1,balanced,0.10086400310198466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,2,power_law_1.2,0.0942463994026184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.2,0.654963207244873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,1,balanced,0.14755200346310934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,2,power_law_1.2,0.1014143943786621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.2,0.7342207908630372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,1,balanced,0.22500266631444296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,2,power_law_1.2,0.10700800418853759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,1,balanced,0.3454879919687907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.2,1.0045120239257812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,1,balanced,0.39930133024851483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,2,power_law_1.2,0.1056704044342041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.2,1.2367487907409669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,1,balanced,0.3903679847717285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,2,power_law_1.2,0.10673919916152955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.2,1.7754432678222656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,1,balanced,0.3874346812566121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,2,power_law_1.2,0.1146239995956421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,1,balanced,0.381440003712972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.2,2.3672319412231446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,2,power_law_1.2,0.12282240390777588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,1,balanced,0.3898880084355672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,2,power_law_1.2,2.9603328704833984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,2,power_law_1.2,0.1297600030899048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,1,balanced,0.385098656018575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,2,power_law_1.2,0.22686080932617186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,1,balanced,0.386405348777771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,2,power_law_1.2,4.830841445922852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,1,balanced,0.3901280164718628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.2,0.2447808027267456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,1,balanced,0.3970346848169963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.2,0.20332798957824708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,2,power_law_1.2,9.137299346923829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.2,0.22299520969390868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,1,balanced,0.4044320185979207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.2,0.26121599674224855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,1,balanced,0.4110720157623291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.2,0.2983423948287964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,1,balanced,0.42159998416900635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.2,0.3631103992462158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,1,balanced,0.4578293164571126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,1,balanced,0.0661653329928716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.2,0.4692224025726318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,1,balanced,0.08132266501585643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.2,0.6492159843444825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,1,balanced,0.4814613262812297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,1,balanced,0.11199999849001567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,1,balanced,0.16410666704177856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.2,0.8433664321899415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,1,balanced,0.24502400557200113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,1,balanced,0.5466346740722656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,2,power_law_1.2,1.0062975883483887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,1,balanced,0.2808159987131755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,1,balanced,0.27480532725652057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,2,power_law_1.2,1.5873984336853026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,1,balanced,0.27295466264088947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,1,balanced,0.6002666552861532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,1,balanced,0.2786986629168193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,2,power_law_1.2,2.8319168090820312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,1,balanced,0.27483733495076496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,1,balanced,0.2761066754659017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,1,balanced,0.28171199560165405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,1,balanced,0.8589066664377848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,1,balanced,0.28676799933115643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,1,balanced,0.2882080078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,1,balanced,0.29814932743708294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,1,balanced,0.30777066946029663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,1,balanced,0.9518346786499023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,1,balanced,0.3190986712773641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,1,balanced,0.3606773217519124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,1,balanced,0.38706668217976886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,1,balanced,0.456826647122701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,1,balanced,1.3642667134602864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,1,balanced,0.5163306792577108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,1,balanced,0.7565653324127197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,1,balanced,1.7728746732076008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,1,balanced,0.8683839639027914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,1,balanced,1.2564000288645427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,1,balanced,1.6337439219156902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,1,balanced,2.2580107053120932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,1,balanced,2.208394686381022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,1,balanced,3.054880142211914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,1,balanced,2.974229176839193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,1,balanced,3.5434773763020835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,1,balanced,3.5441226959228516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,1,balanced,5.6131947835286455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,1,power_law_1.01,0.0762112021446228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,1,power_law_1.01,0.09285759925842285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,1,power_law_1.01,0.1128000020980835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,1,power_law_1.01,0.1674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,1,power_law_1.01,0.202675199508667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,1,balanced,5.639477411905925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,1,power_law_1.01,0.30103681087493894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,1,power_law_1.01,0.32193920612335203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,1,power_law_1.01,0.33346560001373293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,1,power_law_1.01,0.3388736009597778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,1,power_law_1.01,0.06316800117492676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,1,power_law_1.01,0.3458175897598267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,1,power_law_1.01,0.07341439723968506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,1,power_law_1.01,0.3552191972732544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,1,power_law_1.01,0.09207040071487427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,1,power_law_1.01,0.358515191078186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,1,power_law_1.01,0.12348799705505371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,1,power_law_1.01,0.161190402507782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,1,power_law_1.01,0.36540799140930175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,1,power_law_1.01,0.2240959882736206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,1,power_law_1.01,0.3784768104553223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,1,power_law_1.01,0.2337280035018921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,1,power_law_1.01,0.3898751974105835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,1,power_law_1.01,0.24438400268554689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,1,power_law_1.01,0.40298237800598147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,1,power_law_1.01,0.24401919841766356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,1,power_law_1.01,0.4192831993103027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,1,power_law_1.01,0.2425152063369751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,1,power_law_1.01,0.45985918045043944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,1,power_law_1.01,0.24993278980255126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,1,power_law_1.01,0.49379839897155764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,1,power_law_1.01,0.26250879764556884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,1,power_law_1.01,0.5821695804595948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,1,power_law_1.01,0.2678335905075073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,1,power_law_1.01,0.6680704116821289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,1,power_law_1.01,0.2731328010559082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,1,power_law_1.01,0.840998363494873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,1,power_law_1.01,0.2823807954788208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,1,power_law_1.01,0.9960384368896484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,1,power_law_1.01,0.2938240051269531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,1,power_law_1.01,1.3262784004211425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,1,power_law_1.01,0.3118079900741577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,1,power_law_1.01,1.6625728607177734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,1,power_law_1.01,0.3634239912033081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,1,power_law_1.01,0.4001920223236084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,1,power_law_1.01,2.320537567138672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,1,power_law_1.01,0.48531198501586914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,1,power_law_1.01,2.99486083984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,1,power_law_1.01,0.564089584350586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,1,power_law_1.01,3.659436798095703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,1,power_law_1.01,0.7348735809326172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,1,power_law_1.01,0.9039360046386719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,1,power_law_1.01,5.663008117675782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,1,power_law_1.01,1.242080020904541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,1,power_law_1.01,1.5785408020019531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,1,power_law_1.01,11.06595230102539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,1,power_law_1.01,2.2467840194702147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,1,balanced,11.152591705322266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,1,power_law_1.01,2.9230079650878906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,1,power_law_1.01,3.603993606567383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,1,power_law_1.01,5.645497512817383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,1,balanced,11.234891255696615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,1,power_law_1.01,11.118463897705078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,1,power_law_1.2,0.07667840123176575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,1,power_law_1.2,0.0840831995010376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,1,power_law_1.2,0.10997120141983033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,1,power_law_1.2,0.15088000297546386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,1,power_law_1.2,0.0657151997089386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,1,power_law_1.2,0.18698240518569947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,1,power_law_1.2,0.0719488024711609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,1,power_law_1.2,0.3004863977432251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,1,power_law_1.2,0.08880000114440918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,8,balanced,0.06309866905212402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,1,power_law_1.2,0.32665600776672366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,1,power_law_1.2,0.11919360160827637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,8,balanced,0.0711893339951833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,1,power_law_1.2,0.33701119422912595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,1,power_law_1.2,0.13886719942092896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,8,balanced,0.08893866340319316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,8,balanced,0.08904533584912618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,8,balanced,0.13019733627637228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,1,power_law_1.2,0.3443072080612183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,1,power_law_1.2,0.21960320472717285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,8,balanced,0.09529067079226176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,8,balanced,0.20312533775965372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,1,power_law_1.2,0.3527744054794312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,1,power_law_1.2,0.23605759143829347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,8,balanced,0.12954666217168173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,8,balanced,0.3396426836649577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,1,power_law_1.2,0.3663680076599121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,8,balanced,0.19930134216944376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,1,power_law_1.2,0.24535679817199707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,8,balanced,0.44923198223114014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,8,balanced,0.3399306535720825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,1,power_law_1.2,0.3674688100814819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,8,balanced,0.45129601160685223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,1,power_law_1.2,0.2500416040420532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,8,balanced,0.561621348063151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,8,balanced,0.4211680094401042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,1,power_law_1.2,0.3813055992126465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,1,power_law_1.2,0.24982399940490724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,8,balanced,0.8000746568044027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,8,balanced,0.4275413354237874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,1,power_law_1.2,0.38497281074523926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,1,power_law_1.2,0.2604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,8,balanced,0.7958079973856608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,8,balanced,0.42601064840952557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,1,power_law_1.2,0.41075839996337893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,8,balanced,0.8155519962310791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,1,power_law_1.2,0.2695168018341064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,8,balanced,0.43675732612609863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,1,power_law_1.2,0.4165823936462402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,8,balanced,0.8359306653340658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,1,power_law_1.2,0.2790080070495605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,8,balanced,0.45801599820454914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,1,power_law_1.2,0.43140478134155275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,8,balanced,0.8360479672749838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,1,power_law_1.2,0.2834687948226929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,8,balanced,0.42958398660024005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,8,balanced,0.8333439826965332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,1,power_law_1.2,0.4750976085662842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,1,power_law_1.2,0.29966719150543214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,8,balanced,0.46935999393463135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,8,balanced,0.8327999909718832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,1,power_law_1.2,0.5116543769836426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,1,power_law_1.2,0.306329607963562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,8,balanced,0.44437865416208905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,1,power_law_1.2,0.5991615772247314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,1,power_law_1.2,0.32238719463348386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,1,power_law_1.2,0.6789951801300049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,8,balanced,0.45823466777801514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,1,power_law_1.2,0.36720640659332277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,1,power_law_1.2,0.8453503608703613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,8,balanced,0.46449601650238037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,1,power_law_1.2,0.4149631977081299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,1,power_law_1.2,0.9985983848571778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,1,power_law_1.2,0.49628162384033203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,8,balanced,0.48419201374053955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,1,power_law_1.2,1.3326656341552734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,1,power_law_1.2,0.5747903823852539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,1,power_law_1.2,1.6538175582885741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,1,power_law_1.2,0.7484992027282715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,8,balanced,0.4968106746673584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,1,power_law_1.2,2.3350784301757814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,1,power_law_1.2,0.912723159790039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,1,power_law_1.2,2.985728073120117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,1,power_law_1.2,1.2438655853271485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,8,balanced,0.527184009552002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,1,power_law_1.2,3.6351871490478516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,1,power_law_1.2,1.581491184234619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,1,power_law_1.2,5.630259323120117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,1,power_law_1.2,2.2479103088378904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,8,balanced,0.558240016301473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,1,power_law_1.2,2.9239551544189455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,1,power_law_1.2,11.033465576171874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,1,power_law_1.2,3.6051841735839845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,1,power_law_1.2,5.645363235473633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,8,balanced,0.6105386813481649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,1,power_law_1.2,11.100569915771484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,8,balanced,0.7192906538645426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,8,balanced,1.0636320114135742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,8,balanced,1.170522689819336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,8,balanced,1.5766347249348958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,8,balanced,2.139242649078369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,8,balanced,0.06482666730880737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,8,balanced,0.07352533439795177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,8,balanced,3.1808319091796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,8,balanced,0.09455999732017517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,8,balanced,0.15085867047309875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,8,balanced,0.19819200038909912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,8,balanced,0.35254931449890137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,8,balanced,0.4566719929377238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,8,balanced,0.4601653416951497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,8,balanced,0.44966399669647217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,8,balanced,0.41038934389750165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,8,balanced,0.4161173502604167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,8,balanced,0.418720006942749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,8,balanced,0.416378657023112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,8,balanced,0.4262186686197917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,8,balanced,0.44741864999135333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,8,balanced,0.8175199826558431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,8,power_law_1.01,0.20608000755310057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,8,balanced,0.4264959891637166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,8,balanced,0.8265759944915771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,8,power_law_1.01,0.32433280944824217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,8,balanced,0.4437439839045207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,8,balanced,0.797754685084025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,8,power_law_1.01,0.551206398010254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,8,balanced,0.4375786781311035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,8,balanced,0.8240319887797037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,8,power_law_1.01,0.20095360279083252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,8,balanced,0.4728906551996867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,8,balanced,0.8171947002410889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,8,power_law_1.01,0.32069759368896483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,8,power_law_1.01,0.3742271900177002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,8,balanced,0.4541333516438802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,8,balanced,0.8179999987284342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,8,power_law_1.01,0.49891839027404783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,8,balanced,0.4782506624857585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,8,balanced,0.8140052954355875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,8,power_law_1.01,0.570963191986084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,8,power_law_1.01,0.6003903865814209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,8,balanced,0.8403573036193848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,8,balanced,0.5046773354212443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,8,power_law_1.01,0.6643392086029053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,8,power_law_1.01,0.6212543964385986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,8,balanced,0.8410879770914713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,8,power_law_1.01,0.6506688117980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,8,balanced,0.5201599995295206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,8,power_law_1.01,0.6733888149261474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,8,power_law_1.01,0.67259521484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,8,balanced,0.8618613084157308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,8,power_law_1.01,0.7129471778869629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,8,balanced,0.5923306544621786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,8,power_law_1.01,0.6854080200195313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,8,power_law_1.01,0.6956031799316407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,8,power_law_1.01,0.7116479873657227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,8,balanced,0.9444213708241781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,8,power_law_1.01,0.7476480007171631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,8,balanced,0.971887985865275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,8,power_law_1.01,0.8367551803588867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,8,power_law_1.01,0.9020480155944824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,8,power_law_1.01,1.0361791610717774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,8,balanced,1.519584019978841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,8,power_law_1.01,1.226636791229248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,8,balanced,5.63044802347819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,8,power_law_1.01,1.4498047828674316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,8,power_law_1.01,1.7261951446533204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,8,balanced,1.0169066588083904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,8,power_law_1.01,2.3056768417358398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,8,power_law_1.01,2.7433536529541014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,8,balanced,1.4369600613911946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,8,power_law_1.01,3.499135971069336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,8,power_law_1.01,5.082054519653321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,8,power_law_1.01,9.711231994628907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,8,balanced,1.4240214029947917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,8,balanced,2.051232020060221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,8,balanced,1.925066630045573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,8,balanced,2.867706616719564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,8,balanced,2.8267361323038735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,8,balanced,4.174031893412272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,8,balanced,4.94867738087972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,8,power_law_1.01,0.12342400550842285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,8,power_law_1.01,0.19680000543594361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,8,power_law_1.01,0.3172287940979004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,8,power_law_1.01,0.1257024049758911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,8,power_law_1.01,0.17180800437927246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,8,power_law_1.01,0.22894721031188964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,8,power_law_1.01,0.2488192081451416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,8,balanced,7.000133514404297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,8,power_law_1.01,0.32483201026916503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,8,power_law_1.01,0.3286720037460327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,8,power_law_1.01,0.1435263991355896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,8,power_law_1.01,0.33973119258880613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,8,power_law_1.01,0.22710399627685546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,8,power_law_1.01,0.37727999687194824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,8,power_law_1.01,0.3610111951828003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,8,power_law_1.01,0.36463360786437987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,8,power_law_1.01,0.12643840312957763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,8,power_law_1.01,0.3597248077392578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,8,power_law_1.01,0.18035839796066283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,8,power_law_1.01,0.3731071949005127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,8,power_law_1.01,0.26743040084838865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,8,power_law_1.01,0.3713023900985718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,8,power_law_1.01,0.278656005859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,8,power_law_1.01,0.4080959796905518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,8,power_law_1.01,0.3627327919006348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,8,power_law_1.01,0.4080512046813965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,8,power_law_1.01,0.34963839054107665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,8,power_law_1.01,0.43865599632263186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,8,power_law_1.01,0.3386240005493164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,8,power_law_1.01,0.46648321151733396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,8,power_law_1.01,0.3678335905075073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,8,power_law_1.01,0.4888256072998047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,8,power_law_1.01,0.35612800121307375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,8,power_law_1.01,0.5473792076110839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,8,power_law_1.01,0.37360639572143556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,8,power_law_1.01,0.6545663833618164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,8,power_law_1.01,0.39214720726013186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,8,power_law_1.01,0.7624832153320312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,8,power_law_1.01,0.40819201469421384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,8,power_law_1.01,0.4202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,8,power_law_1.01,0.9559679985046386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,8,power_law_1.01,1.1858559608459474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,8,power_law_1.01,0.41597437858581543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,8,power_law_1.01,0.419539213180542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,8,power_law_1.01,1.6645183563232422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,8,power_law_1.01,0.4601151943206787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,8,power_law_1.01,2.0531200408935546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,8,power_law_1.01,0.4886911869049072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,8,power_law_1.01,2.482809638977051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,8,power_law_1.01,0.533247995376587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,8,power_law_1.01,3.6283710479736326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,8,power_law_1.01,0.635923194885254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,8,power_law_1.01,0.7450175762176514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,8,power_law_1.01,7.128518676757812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,8,power_law_1.01,0.9325695991516113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,8,power_law_1.01,1.1644160270690918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,8,power_law_1.01,1.4935104370117187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,8,power_law_1.01,1.9889408111572267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,8,power_law_1.01,2.34911994934082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,8,power_law_1.01,3.2281982421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,8,power_law_1.01,6.2096702575683596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,8,power_law_1.2,0.20595839023590087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,8,power_law_1.2,0.3234944105148315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,8,power_law_1.2,0.3399168014526367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,8,power_law_1.2,0.1789888024330139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,8,power_law_1.2,0.24816639423370362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,8,power_law_1.2,0.35819520950317385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,8,power_law_1.2,0.4216896057128906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,8,power_law_1.2,0.5435135841369629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,8,power_law_1.2,0.5595327854156494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,8,power_law_1.2,0.5525631904602051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,8,power_law_1.2,0.6132415771484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,8,power_law_1.2,0.6377088069915772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,8,power_law_1.2,0.6439551830291748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,8,power_law_1.2,0.6536448001861572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,8,power_law_1.2,0.667628812789917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,8,power_law_1.2,0.6658815860748291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,8,power_law_1.2,0.7047872066497802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,8,power_law_1.2,0.7524288177490235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,8,power_law_1.2,0.7901887893676758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,8,power_law_1.2,0.13671679496765138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,8,power_law_1.2,0.8411968231201172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,8,power_law_1.2,0.21517438888549806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,8,power_law_1.2,1.001471996307373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,8,power_law_1.2,0.17906559705734254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,8,power_law_1.2,1.1402432441711425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,8,power_law_1.2,0.12041599750518799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,8,power_law_1.2,1.3280896186828612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,8,power_law_1.2,0.18986239433288574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,8,power_law_1.2,0.22532479763031005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,8,power_law_1.2,1.7439615249633789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,8,power_law_1.2,0.2405247926712036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,8,power_law_1.2,1.8880512237548828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,8,power_law_1.2,0.3280384063720703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,8,power_law_1.2,2.52410888671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,8,power_law_1.2,0.3601279973983765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,8,power_law_1.2,2.982028770446777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,8,power_law_1.2,0.3231168031692505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,8,power_law_1.2,3.989459228515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,8,power_law_1.2,0.35146241188049315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,8,power_law_1.2,5.8862464904785154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,8,power_law_1.2,0.37292160987854006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,8,power_law_1.2,0.36618239879608155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,8,power_law_1.2,11.402848052978516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,8,power_law_1.2,0.3677504062652588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,8,power_law_1.2,0.4098368167877197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,8,power_law_1.2,0.39687678813934324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,8,power_law_1.2,0.3990400075912476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,8,power_law_1.2,0.45864319801330566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,8,power_law_1.2,0.4906047821044922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,8,power_law_1.2,0.5376255989074707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,8,power_law_1.2,0.5961023807525635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,8,power_law_1.2,0.7369472026824951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,8,power_law_1.2,0.8673343658447266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,8,power_law_1.2,1.109273624420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,8,power_law_1.2,1.3387264251708983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,8,power_law_1.2,1.794438362121582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,8,power_law_1.2,0.14394880533218385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,8,power_law_1.2,2.2503807067871096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,8,power_law_1.2,0.21848320960998535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,8,power_law_1.2,2.6857919692993164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,4,balanced,0.07866133252779643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,8,power_law_1.2,0.18507519960403443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,4,balanced,0.09705600142478943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,4,balanced,0.1183786690235138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,8,power_law_1.2,4.205996704101563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,8,power_law_1.2,0.13155200481414794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,4,balanced,0.19412267208099365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,4,balanced,0.32771732409795123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,8,power_law_1.2,0.1855232000350952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,8,power_law_1.2,7.9925697326660154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,4,balanced,0.528714656829834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,8,power_law_1.2,0.20926079750061036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,4,balanced,0.7657546997070312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,4,balanced,0.7789546648661295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,8,power_law_1.2,0.2601151943206787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,4,balanced,0.7629120349884033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,8,power_law_1.2,0.32721920013427735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,4,balanced,0.7605813344319662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,4,balanced,0.8004426956176758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,8,power_law_1.2,0.3157952070236206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,4,balanced,0.782906691233317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,8,power_law_1.2,0.32849280834197997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,4,balanced,0.791429360707601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,4,balanced,0.7952586809794108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,8,power_law_1.2,0.3375296115875244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,4,balanced,0.7688852945963541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,8,power_law_1.2,0.34029440879821776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,4,balanced,0.7958986759185791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,8,power_law_1.2,0.37867519855499265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,4,balanced,0.7967360019683838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,8,power_law_1.2,0.36278400421142576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,4,balanced,0.7894666989644369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,8,power_law_1.2,0.39149439334869385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,4,balanced,0.8152853647867838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,8,power_law_1.2,0.3930752038955688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,8,power_law_1.2,0.4147647857666016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,4,balanced,0.819648027420044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,8,power_law_1.2,0.43704957962036134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,8,power_law_1.2,0.4650047779083252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,4,balanced,0.8424213727315267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,8,power_law_1.2,0.5256639957427979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,8,power_law_1.2,0.5807295799255371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,4,balanced,0.8698399861653646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,8,power_law_1.2,0.6915520191192627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,8,power_law_1.2,0.8044544219970703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,8,power_law_1.2,0.9997119903564453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,4,balanced,0.9130187034606934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,8,power_law_1.2,1.238912010192871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,8,power_law_1.2,1.6602752685546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,8,power_law_1.2,2.1071487426757813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,4,balanced,1.0251893202463787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,8,power_law_1.2,2.4593664169311524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,8,power_law_1.2,3.50145263671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,8,power_law_1.2,7.546041870117188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,4,balanced,1.563253402709961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,4,balanced,1.581925392150879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,4,balanced,2.1807467142740884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,4,balanced,0.06010133524735769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,4,balanced,0.06945066650708516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,4,balanced,0.08718933661778767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,4,balanced,0.1283093293507894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,4,balanced,0.20312533775965372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,4,balanced,0.3178559939066569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,4,balanced,0.4484693209330241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,4,balanced,0.4373386700948079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,4,balanced,0.42999998728434247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,4,balanced,0.4238080183664958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,4,balanced,0.42209064960479736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,4,balanced,3.0164639155069985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,4,balanced,0.43341867129007977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,4,balanced,0.4206560055414836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,4,balanced,0.4277760187784831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,4,balanced,0.44706134001413983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,4,balanced,0.45418667793273926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,4,balanced,0.4580586751302083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,4,balanced,0.47042667865753174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,4,balanced,0.4742026726404826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,4,balanced,0.5100213289260864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,4,balanced,0.5412693421045939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,4,balanced,0.5995093186696371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,4,balanced,0.6672373612721761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,4,balanced,4.393952051798503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,4,balanced,0.06878933310508728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,4,balanced,0.08369599779446919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,4,balanced,0.10874666770299275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,4,balanced,0.15292267004648843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,4,balanced,0.8174346288045248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,4,balanced,0.22522666056950888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,4,balanced,0.36030932267506915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,4,balanced,0.48764801025390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,4,balanced,0.4573386510213216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,4,balanced,0.442906657854716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,4,balanced,0.4327733516693115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,4,balanced,1.1410826841990154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,4,balanced,0.4299360116322835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,4,balanced,0.4270613193511963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,4,balanced,0.42235199610392254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,4,balanced,0.4324800173441569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,4,balanced,0.43756266434987384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,4,balanced,0.43091734250386554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,4,balanced,0.43245867888132733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,4,balanced,0.4583359956741333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,4,balanced,1.3663093249003093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,4,balanced,0.46138668060302734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,4,balanced,0.48977065086364746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,4,balanced,0.5004266500473022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,4,balanced,0.5563679933547974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,4,balanced,1.8163305918375652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,4,balanced,0.5868533452351888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,4,power_law_1.01,0.12329599857330323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,4,power_law_1.01,0.19035520553588867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,4,power_law_1.01,0.30663039684295657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,4,balanced,7.711130777994792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,4,power_law_1.01,0.1646783947944641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,4,balanced,0.7231840292612711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,4,power_law_1.01,0.2568831920623779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,4,balanced,2.368410587310791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,4,power_law_1.01,0.3379008054733276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,4,power_law_1.01,0.4088831901550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,4,power_law_1.01,0.503001594543457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,4,power_law_1.01,0.5438464164733887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,4,power_law_1.01,0.5484416007995605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,4,balanced,1.0859253406524658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,4,power_law_1.01,0.5829440116882324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,4,power_law_1.01,0.6051136016845703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,4,power_law_1.01,0.5827968120574951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,4,power_law_1.01,0.6296127796173095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,4,power_law_1.01,0.6713088035583497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,4,power_law_1.01,0.6420608043670655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,4,power_law_1.01,0.6763264179229737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,4,power_law_1.01,0.6796160221099854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,4,power_law_1.01,0.7009471893310547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,4,balanced,1.2338133653004963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,4,power_law_1.01,0.8095616340637207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,4,balanced,3.5512161254882812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,4,power_law_1.01,0.8510080337524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,4,power_law_1.01,0.9795392036437989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,4,power_law_1.01,1.1270591735839843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,4,power_law_1.01,1.385209560394287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,4,power_law_1.01,1.6592512130737305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,4,power_law_1.01,2.178783988952637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,4,power_law_1.01,2.7753984451293947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,4,power_law_1.01,3.201900863647461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,4,power_law_1.01,4.994489669799805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,4,balanced,1.6887733141581218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,4,power_law_1.01,8.641414642333984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,4,balanced,2.2027146021525064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,4,balanced,6.591786702473958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,4,balanced,3.3174079259236655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,4,power_law_1.01,0.09404799938201905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,4,power_law_1.01,0.13368959426879884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,4,power_law_1.01,0.205132794380188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,4,power_law_1.01,0.11650559902191163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,4,power_law_1.01,0.17034239768981935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,4,power_law_1.01,0.21231999397277831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,4,power_law_1.01,0.26019198894500734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,4,power_law_1.01,0.3165503978729248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,4,power_law_1.01,0.3175935983657837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,4,power_law_1.01,0.3297663927078247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,4,power_law_1.01,0.34754559993743894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,4,power_law_1.01,0.34858880043029783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,4,power_law_1.01,0.3622591972351074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,4,power_law_1.01,0.37109758853912356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,4,power_law_1.01,0.38366720676422117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,4,power_law_1.01,0.3892352104187012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,4,power_law_1.01,0.3958143949508667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,4,power_law_1.01,0.41403517723083494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,4,power_law_1.01,0.44531841278076173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,4,power_law_1.01,0.5141183853149414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,4,power_law_1.01,0.5733248233795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,4,power_law_1.01,0.6863743782043457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,4,power_law_1.01,0.7751999855041504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,4,power_law_1.01,1.0728639602661132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,4,power_law_1.01,0.11130880117416382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,4,power_law_1.01,1.2385279655456543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,4,power_law_1.01,0.1486143946647644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,4,power_law_1.01,1.8455743789672852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,4,power_law_1.01,0.22445440292358398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,4,power_law_1.01,2.148896026611328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,4,power_law_1.01,0.12234879732131958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,4,power_law_1.01,2.7273279190063477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,4,balanced,6.037984212239583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,4,power_law_1.01,0.17060480117797852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,4,power_law_1.01,3.7307903289794924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,4,power_law_1.01,0.23571839332580566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,4,power_law_1.01,0.2500864028930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,4,power_law_1.01,7.7135871887207035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,4,power_law_1.01,0.3249919891357422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,4,power_law_1.01,0.33552000522613523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,4,power_law_1.01,0.330348801612854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,4,power_law_1.01,0.3480128049850464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,4,power_law_1.01,0.36389119625091554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,4,power_law_1.01,0.37015039920806886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,4,power_law_1.01,0.3814143896102905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,4,power_law_1.01,0.3786367893218994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,4,power_law_1.2,0.12126079797744752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,4,power_law_1.01,0.38962559700012206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,4,power_law_1.2,0.18913279771804808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,4,power_law_1.01,0.4002431869506836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,4,power_law_1.2,0.17132799625396727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,4,power_law_1.01,0.42133121490478515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,4,power_law_1.2,0.16272000074386597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,4,power_law_1.01,0.45172481536865233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,4,power_law_1.2,0.24086399078369142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,4,power_law_1.01,0.4971776008605957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,4,power_law_1.01,0.557478380203247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,4,power_law_1.2,0.3358272075653076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,4,power_law_1.2,0.4028736114501953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,4,power_law_1.01,0.6296576023101806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,4,power_law_1.2,0.5011712074279785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,4,power_law_1.01,0.7620800018310547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,4,power_law_1.2,0.5395071983337403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,4,power_law_1.01,1.0144384384155274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,4,power_law_1.2,0.540172815322876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,4,power_law_1.01,1.180742359161377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,4,power_law_1.2,0.5754816055297851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,4,power_law_1.01,1.5820992469787598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,4,power_law_1.2,0.5946176052093506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,4,power_law_1.01,2.108563232421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,4,power_law_1.2,0.5875391960144043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,4,power_law_1.01,2.4265151977539063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,4,power_law_1.2,0.6418816089630127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,4,power_law_1.01,3.6430591583251952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,4,power_law_1.2,0.6607615947723389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,4,power_law_1.01,6.912000274658203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,4,power_law_1.2,0.6593023777008057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,4,power_law_1.2,0.6718783855438233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,4,power_law_1.2,0.7238399982452393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,4,power_law_1.2,0.7703360080718994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,4,power_law_1.2,0.8836735725402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,4,power_law_1.2,0.9445952415466309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,4,power_law_1.2,1.1410752296447755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,4,power_law_1.2,1.229593563079834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,4,power_law_1.2,1.5852671623229981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,4,power_law_1.2,1.7423871994018554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,4,power_law_1.2,2.455232048034668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,4,power_law_1.2,2.8114879608154295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,4,power_law_1.2,3.6007232666015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,4,power_law_1.2,5.314137649536133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,4,power_law_1.2,10.012863922119141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,4,power_law_1.2,0.11114239692687988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,4,power_law_1.2,0.14907519817352294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,4,power_law_1.2,0.09217280149459839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,4,power_law_1.2,0.1455296039581299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,4,power_law_1.2,0.12857600450515747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,4,power_law_1.2,0.1199295997619629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,4,power_law_1.2,0.11399680376052856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,4,power_law_1.2,0.17656320333480835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,4,power_law_1.2,0.11251200437545776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,4,power_law_1.2,0.21397120952606202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,4,power_law_1.2,0.1673087954521179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,4,power_law_1.2,0.2388416051864624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,4,power_law_1.2,0.20937600135803222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,4,power_law_1.2,0.31255040168762205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,4,power_law_1.2,0.23594241142272948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,4,power_law_1.2,0.3198911905288696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,4,power_law_1.2,0.2861824035644531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,4,power_law_1.2,0.3540992021560669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,4,power_law_1.2,0.3162879943847656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,4,power_law_1.2,0.3502784013748169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,4,power_law_1.2,0.32476799488067626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,4,power_law_1.2,0.3554111957550049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,4,power_law_1.2,0.3736511945724487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,4,power_law_1.2,0.32074239253997805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,4,power_law_1.2,0.3868799924850464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,4,power_law_1.2,0.35535359382629395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,4,power_law_1.2,0.39874560832977296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,4,power_law_1.2,0.36135680675506593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,4,power_law_1.2,0.3602880001068115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,4,power_law_1.2,0.39507839679718015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,4,power_law_1.2,0.37112960815429685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,4,power_law_1.2,0.40921602249145506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,4,power_law_1.2,0.3852992057800293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,4,power_law_1.2,0.4369215965270996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,4,power_law_1.2,0.3991039991378784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,4,power_law_1.2,0.4780992031097412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,4,power_law_1.2,0.44101758003234864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,4,power_law_1.2,0.5439167976379394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,4,power_law_1.2,0.46741762161254885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,4,power_law_1.2,0.60316162109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,4,power_law_1.2,0.5505919933319092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,4,power_law_1.2,0.7229055881500244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,4,power_law_1.2,0.5998464107513428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,4,power_law_1.2,0.8237119674682617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,4,power_law_1.2,0.734335994720459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,4,power_law_1.2,0.9981311798095703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,4,power_law_1.2,0.844057559967041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,4,power_law_1.2,1.333625602722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,2,balanced,0.0781173308690389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,2,balanced,0.09563733140627544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,4,power_law_1.2,1.1157823562622071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,4,power_law_1.2,1.6630016326904298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,2,balanced,0.11879466970761617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,4,power_law_1.2,1.3970879554748534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,2,balanced,0.19115199645360312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,4,power_law_1.2,2.1375808715820312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,2,balanced,0.3172159989674886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,4,power_law_1.2,1.8137983322143554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,2,balanced,0.5321546792984009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,4,power_law_1.2,2.719878387451172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,2,balanced,0.750058650970459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,4,power_law_1.2,2.274239921569824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,2,balanced,0.7515520254770914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,4,power_law_1.2,4.265977478027343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,2,balanced,0.7443679968516032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,4,power_law_1.2,2.6178688049316405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,2,balanced,0.747770627339681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,4,power_law_1.2,7.371603393554688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,4,power_law_1.2,4.012921524047852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,2,balanced,0.753216028213501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,2,balanced,0.7497599919637045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,2,balanced,0.734117348988851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,4,power_law_1.2,7.620953369140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,2,balanced,0.744271993637085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,2,balanced,0.7603946526845297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,2,balanced,0.7586987018585205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,2,balanced,0.768069346745809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,2,balanced,0.7806453704833984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,2,balanced,0.06609599788983662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,2,balanced,0.07634133100509644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,2,balanced,0.7965866724650065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,2,balanced,0.09322133660316467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,2,balanced,0.13805333773295084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,2,balanced,0.8397546609242758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,2,balanced,0.2044586737950643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,2,balanced,0.319050669670105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,2,balanced,0.44594132900238037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,2,balanced,0.43983999888102215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,2,balanced,0.879418690999349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,2,balanced,0.436570684115092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,2,balanced,0.4334346850713094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,2,balanced,0.440229336420695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,2,balanced,0.44020267327626544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,2,balanced,0.9591680367787679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,2,balanced,0.44786131381988525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,2,balanced,0.4485866626103719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,2,balanced,0.4630613327026367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,2,balanced,0.46163733800252277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,2,balanced,1.0414559841156006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,2,balanced,0.48026665051778156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,2,balanced,0.5018399953842163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,2,balanced,0.5219306548436483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,2,balanced,1.2212106386820476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,2,balanced,0.5718079805374146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,2,balanced,0.6131733258565267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,2,balanced,1.6767466862996419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,2,balanced,0.716426690419515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,2,balanced,0.8115306695302328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,2,balanced,1.9538453420003254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,2,balanced,1.0190400282541912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,2,balanced,1.413002649943034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,2,balanced,2.5938933690389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,2,balanced,1.7736266454060872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,2,balanced,3.3440478642781577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,2,balanced,2.3624480565389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,2,balanced,2.9759785334269204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,2,balanced,5.0199893315633135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,2,balanced,0.08112533390522003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,2,power_law_1.01,0.09669119715690613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,2,balanced,0.09572800000508626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,2,balanced,0.11512000362078349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,2,power_law_1.01,0.12645119428634644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,2,balanced,0.1704053282737732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,2,balanced,0.2606559991836548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,2,power_law_1.01,0.19352960586547852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,2,balanced,0.39158399899800617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,2,power_law_1.01,0.15978239774703978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,2,balanced,0.5370773474375407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,2,balanced,0.5202186504999796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,2,power_law_1.01,0.25737600326538085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,2,balanced,4.523189226786296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,2,balanced,0.5069493452707926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,2,power_law_1.01,0.35703039169311523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,2,balanced,0.5066346724828085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,2,balanced,0.502618670463562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,2,power_law_1.01,0.3928447961807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,2,balanced,0.4893706639607747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,2,power_law_1.01,0.4949632167816162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,2,balanced,0.4824906587600708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,2,power_law_1.01,0.517190408706665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,2,balanced,0.48452266057332355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,2,balanced,0.5198986530303955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,2,power_law_1.01,0.522540807723999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,2,balanced,0.5076853434244791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,2,power_law_1.01,0.5728320121765137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,2,balanced,0.5228213469187418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,2,power_law_1.01,0.5856383800506592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,2,balanced,0.5362666845321655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,2,power_law_1.01,0.5953152179718018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,2,power_law_1.01,0.6087744235992432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,2,balanced,0.5507200161616007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,2,power_law_1.01,0.6507967948913574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,2,balanced,0.6010773181915283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,2,power_law_1.01,0.6439871788024902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,2,power_law_1.01,0.6706943988800049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,2,balanced,0.6361120144526163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,2,power_law_1.01,0.7065023899078369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,2,power_law_1.01,0.7305920124053955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,2,balanced,0.7079946994781494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,2,power_law_1.01,0.8025471687316894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,2,power_law_1.01,0.8684543609619141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,2,power_law_1.01,1.032972812652588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,2,balanced,0.783893346786499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,2,power_law_1.01,1.1861056327819823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,2,balanced,9.352906545003256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,2,power_law_1.01,1.452787208557129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,2,power_law_1.01,1.7776384353637695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,2,balanced,0.9711893399556478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,2,power_law_1.01,2.406687927246094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,2,power_law_1.01,2.9409727096557616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,2,power_law_1.01,3.556288146972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,2,balanced,1.4432479540507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,2,power_law_1.01,5.414227294921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,2,balanced,8.572538375854492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,2,power_law_1.01,10.074156951904296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,2,balanced,1.7276585896809895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,2,balanced,2.3436692555745444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,2,balanced,3.000650723775228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,2,balanced,4.523440043131511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,2,power_law_1.01,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,2,power_law_1.01,0.09466879963874816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,2,power_law_1.01,0.13559039831161498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,2,power_law_1.01,0.11105920076370239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,2,power_law_1.01,0.16161919832229615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,2,power_law_1.01,0.22366719245910643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,2,power_law_1.01,0.24968318939208983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,2,power_law_1.01,0.09583359956741333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,2,power_law_1.01,0.2924607992172241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,2,power_law_1.01,0.11622400283813476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,2,power_law_1.01,0.31427199840545655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,2,power_law_1.01,0.16482559442520142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,2,power_law_1.01,0.328601598739624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,2,power_law_1.01,0.3417088031768799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,2,power_law_1.01,0.1387071967124939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,2,balanced,8.60258674621582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,2,power_law_1.01,0.35855998992919924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,2,power_law_1.01,0.1993280053138733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,2,power_law_1.01,0.3631616115570068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,2,power_law_1.01,0.2644032001495361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,2,power_law_1.01,0.3618560075759888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,2,power_law_1.2,0.09580159783363343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,2,power_law_1.01,0.304966402053833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,2,power_law_1.01,0.39175679683685305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,2,power_law_1.2,0.12560640573501586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,2,power_law_1.01,0.36272640228271485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,2,power_law_1.01,0.3914880037307739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,2,power_law_1.2,0.11861120462417603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,2,power_law_1.01,0.3776063919067383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,2,power_law_1.2,0.16147840023040771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,2,power_law_1.01,0.4082304000854492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,2,power_law_1.01,0.38190081119537356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,2,power_law_1.2,0.2277695894241333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,2,power_law_1.01,0.4447296142578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,2,power_law_1.01,0.3905728101730347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,2,power_law_1.2,0.31693439483642577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,2,power_law_1.01,0.48314881324768066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,2,power_law_1.01,0.4034431934356689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,2,power_law_1.01,0.5460608005523682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,2,power_law_1.2,0.34168319702148436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,2,power_law_1.01,0.40919041633605957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,2,power_law_1.01,0.6091328144073487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,2,power_law_1.2,0.4704768180847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,2,power_law_1.01,0.42381439208984373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,2,power_law_1.01,0.751423978805542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,2,power_law_1.2,0.5119359970092774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,2,power_law_1.01,0.43872637748718263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,2,power_law_1.01,0.8800576210021973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,2,power_law_1.2,0.5210432052612305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,2,power_law_1.01,0.4301504135131836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,2,power_law_1.2,0.5397632122039795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,2,power_law_1.01,0.4604032039642334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,2,power_law_1.01,1.1614591598510742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,2,power_law_1.2,0.5442624092102051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,2,power_law_1.01,0.4808640003204346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,2,power_law_1.01,1.430668830871582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,2,power_law_1.2,0.5980735778808594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,2,power_law_1.01,0.5165056228637696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,2,power_law_1.01,1.9809087753295898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,2,power_law_1.2,0.6096960067749023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,2,power_law_1.01,0.580134391784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,2,power_law_1.01,2.4426816940307616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,2,power_law_1.2,0.6241600036621093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,2,power_law_1.01,0.6500736236572265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,2,power_law_1.01,3.0304000854492186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,2,power_law_1.2,0.6419904232025146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,2,power_law_1.01,0.7767551898956299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,2,power_law_1.01,4.679103851318359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,2,power_law_1.2,0.6448383808135987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,2,power_law_1.01,0.9052224159240723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,2,power_law_1.2,0.7163712024688721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,2,power_law_1.01,1.1881152153015138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,2,power_law_1.01,8.752178955078126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,2,power_law_1.2,0.7608511924743653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,2,power_law_1.01,1.4420736312866211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,2,power_law_1.2,0.8677248001098633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,2,power_law_1.01,2.0453887939453126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,2,power_law_1.2,0.9255167961120605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,2,power_law_1.01,2.504268836975098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,2,power_law_1.2,1.0960512161254883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,2,power_law_1.01,3.1748672485351563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,2,power_law_1.2,1.247059154510498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,2,power_law_1.2,1.6184255599975585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,2,power_law_1.01,4.602963256835937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,2,power_law_1.2,1.8963136672973633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,2,power_law_1.01,9.099359893798828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,2,power_law_1.2,2.5075199127197267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,2,power_law_1.2,3.2375999450683595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,2,power_law_1.2,3.6904575347900392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,2,power_law_1.2,5.416889572143555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,2,power_law_1.2,10.045830535888673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,2,power_law_1.2,0.07418879866600037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,2,power_law_1.2,0.09837440252304078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,2,power_law_1.2,0.09848319888114929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,2,power_law_1.2,0.1105728030204773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,2,power_law_1.2,0.1551424026489258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,2,power_law_1.2,0.19064320325851442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,2,power_law_1.2,0.2290816068649292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,2,power_law_1.2,0.2914367914199829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,2,power_law_1.2,0.3150207996368408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,2,power_law_1.2,0.32707839012145995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,2,power_law_1.2,0.31734399795532225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,2,power_law_1.2,0.3470911979675293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,2,power_law_1.2,0.35954558849334717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,2,power_law_1.2,0.3719167947769165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,2,power_law_1.2,0.3894912004470825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,2,power_law_1.2,0.40114560127258303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,2,power_law_1.2,0.4207615852355957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,2,power_law_1.2,0.46199679374694824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,2,power_law_1.2,0.09228799939155578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,2,power_law_1.2,0.4939455986022949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,2,power_law_1.2,0.11804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,2,power_law_1.2,0.5712255954742431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,2,power_law_1.2,0.1142848014831543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,2,power_law_1.2,0.6445888042449951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,2,power_law_1.2,0.1439039945602417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,2,power_law_1.2,0.7842048168182373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,2,power_law_1.2,0.18644479513168336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,2,power_law_1.2,0.9327872276306153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,2,power_law_1.2,0.23017599582672119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,2,power_law_1.2,1.1896256446838378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,2,power_law_1.2,0.26620800495147706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,2,power_law_1.2,1.4652799606323241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,2,power_law_1.2,0.35105280876159667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,2,power_law_1.2,1.9854528427124023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,2,power_law_1.2,0.353222393989563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,2,power_law_1.2,2.529542350769043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,2,power_law_1.2,0.3688640117645264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,1,balanced,0.09155733386675517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,2,power_law_1.2,3.1054719924926757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,1,balanced,0.10847466190656026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,2,power_law_1.2,0.37792000770568845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,1,balanced,0.13332800070444742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,2,power_law_1.2,4.958726501464843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,2,power_law_1.2,0.39457919597625735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,1,balanced,0.21099199851353964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,1,balanced,0.3386506636937459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,2,power_law_1.2,0.4022848129272461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,1,balanced,0.5715413490931193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,2,power_law_1.2,9.397293090820312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,2,power_law_1.2,0.4233407974243164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,1,balanced,0.8051253159840902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,1,balanced,0.8059413433074951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,2,power_law_1.2,0.4303296089172363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,1,balanced,0.8039893309275309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,2,power_law_1.2,0.44167680740356446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,1,balanced,0.8039093017578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,1,balanced,0.8101279735565186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,1,balanced,0.07524266839027405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,2,power_law_1.2,0.4593855857849121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,1,balanced,0.0897813340028127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,1,balanced,0.8116160233815511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,2,power_law_1.2,0.5049920082092285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,1,balanced,0.11501333117485046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,1,balanced,0.8006772994995117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,1,balanced,0.15684266885121664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,2,power_law_1.2,0.5287424087524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,1,balanced,0.8168213367462158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,1,balanced,0.2541973392168681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,2,power_law_1.2,0.6105023860931397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,1,balanced,0.8139786720275879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,1,balanced,0.38784531752268475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,1,balanced,0.5510239998499552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,1,balanced,0.8168959617614746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,2,power_law_1.2,0.6968319892883301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,1,balanced,0.5414239962895712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,1,balanced,0.830725351969401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,2,power_law_1.2,0.8048255920410157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,1,balanced,0.5475733280181885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,1,balanced,0.5435786644617716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,1,balanced,0.8736266295115153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,2,power_law_1.2,0.9457344055175781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,1,balanced,0.5458720127741495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,1,balanced,0.9162773291269938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,2,power_law_1.2,1.293785572052002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,1,balanced,0.5521493355433146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,1,balanced,0.552016019821167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,2,power_law_1.2,1.5278207778930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,1,balanced,0.9835999806722006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,1,balanced,0.5549813508987427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,2,power_law_1.2,2.069830322265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,1,balanced,0.5718239943186442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,1,balanced,1.0419092973073323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,1,balanced,0.5765173435211182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,2,power_law_1.2,2.6846080780029298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,1,balanced,0.5877813498179117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,2,power_law_1.2,3.197478485107422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,1,balanced,1.1847253640492756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,1,balanced,0.6372586488723755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,2,power_law_1.2,4.7023872375488285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,1,balanced,0.6706346670786539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,2,power_law_1.2,9.22507553100586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,1,balanced,1.3200426896413167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,1,balanced,0.7518080075581869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,1,balanced,0.8202613194783529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,1,balanced,1.6123199462890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,1,balanced,0.9746987024943033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,1,balanced,2.307472070058187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,1,balanced,1.1296532948811848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,1,balanced,1.4539999961853027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,1,balanced,2.907343864440918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,1,balanced,2.0139573415120444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,1,balanced,3.8048105239868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,1,balanced,2.6508000691731772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,1,balanced,4.766730626424153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,1,balanced,3.515962600708008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,1,balanced,4.363231976826985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,1,balanced,7.111866633097331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,1,balanced,0.11405866344769795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,1,balanced,0.12478933731714885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,1,balanced,0.15530133247375488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,1,balanced,0.21733866135279337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,1,balanced,0.3126399914423625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,1,balanced,0.5035253365834554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,1,balanced,0.6908160050710043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,1,balanced,0.678330659866333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,1,balanced,0.6723413467407227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,1,balanced,0.6678240299224854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,1,balanced,0.6656746864318848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,1,balanced,6.682090759277344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,1,balanced,0.6787947018941244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,1,balanced,0.677296002705892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,1,balanced,0.686240037282308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,1,balanced,0.7083840370178223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,1,balanced,0.6970880031585693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,1,balanced,0.7230239709218343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,1,power_law_1.01,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,1,balanced,0.7599946657816569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,1,power_law_1.01,0.10303360223770142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,1,balanced,0.7932853698730469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,1,power_law_1.01,0.1291264057159424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,1,power_law_1.01,0.1711680054664612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,1,balanced,0.8608373006184896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,1,power_law_1.01,0.26282238960266113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,1,power_law_1.01,0.33897600173950193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,1,balanced,0.9226133028666178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,1,power_law_1.01,0.3991231918334961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,1,power_law_1.01,0.4927680015563965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,1,balanced,1.0639306704203289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,1,power_law_1.01,0.5307456016540527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,1,power_law_1.01,0.559827184677124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,1,power_law_1.01,0.578707218170166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,1,balanced,1.2093653678894043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,1,power_law_1.01,0.5970880031585694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,1,balanced,13.231887817382812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,1,power_law_1.01,0.6138751983642579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,1,power_law_1.01,0.6476736068725586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,1,power_law_1.01,0.648089599609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,1,power_law_1.01,0.6620287895202637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,1,balanced,1.5178613662719727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,1,power_law_1.01,0.702681589126587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,1,power_law_1.01,0.7652287960052491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,1,power_law_1.01,0.7955455780029297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,1,power_law_1.01,0.9018495559692383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,1,power_law_1.01,0.9916223526000977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,1,balanced,2.2135839462280273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,1,power_law_1.01,1.1855551719665527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,1,power_law_1.01,1.3770048141479492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,1,power_law_1.01,1.7437759399414063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,1,balanced,12.701471964518229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,1,power_law_1.01,2.161260795593262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,1,power_law_1.01,2.931679916381836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,1,power_law_1.01,3.705478286743164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,1,balanced,2.780501365661621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,1,power_law_1.01,4.499135971069336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,1,power_law_1.01,6.809740447998047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,1,power_law_1.01,13.05944366455078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,1,balanced,3.747786521911621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,1,balanced,4.7518666585286455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,1,balanced,7.282623926798503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,1,power_law_1.01,0.07088639736175537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,1,power_law_1.01,0.08279039859771728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,1,power_law_1.01,0.10762239694595337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,1,power_law_1.01,0.12951040267944336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,1,power_law_1.01,0.201964807510376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,1,power_law_1.01,0.2661312103271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,1,power_law_1.01,0.2944511890411377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,1,power_law_1.01,0.35389440059661864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,1,power_law_1.01,0.37893118858337405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,1,power_law_1.01,0.3897279977798462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,1,power_law_1.01,0.4121088027954102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,1,power_law_1.01,0.4346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,1,power_law_1.01,0.4291840076446533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,1,power_law_1.01,0.4394239902496338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,1,power_law_1.01,0.45902080535888673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,1,power_law_1.01,0.4863423824310303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,1,power_law_1.01,0.5054463863372802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,1,power_law_1.01,0.5594687938690186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,1,power_law_1.01,0.5993087768554688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,1,power_law_1.01,0.10487040281295776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,1,power_law_1.01,0.7065343856811523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,1,power_law_1.01,0.11936000585556031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,1,power_law_1.01,0.8011263847351074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,1,power_law_1.01,0.14740480184555055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,1,power_law_1.01,0.9932736396789551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,1,power_law_1.01,0.16800639629364014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,1,power_law_1.01,1.1863679885864258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,1,power_law_1.2,0.08349440097808838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,1,power_law_1.01,0.25867519378662107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,1,power_law_1.01,1.5721407890319825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,1,power_law_1.2,0.10228480100631714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,1,power_law_1.01,0.3272511959075928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,1,power_law_1.2,0.1063040018081665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,1,power_law_1.01,1.9584320068359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,1,power_law_1.01,0.3731584072113037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,1,balanced,13.822191874186197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,1,power_law_1.2,0.16628479957580566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,1,power_law_1.01,0.46230401992797854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,1,power_law_1.01,2.7212671279907226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,1,power_law_1.2,0.24097919464111328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,1,power_law_1.01,0.4865600109100342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,1,power_law_1.01,3.497817611694336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,1,power_law_1.2,0.30558080673217775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,1,power_law_1.01,0.4923391819000244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,1,power_law_1.01,4.259820938110352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,1,power_law_1.2,0.34908161163330076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,1,power_law_1.01,0.520633602142334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,1,power_law_1.2,0.47633919715881345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,1,power_law_1.01,6.565164947509766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,1,power_law_1.01,0.5392191886901856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,1,power_law_1.2,0.530188798904419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,1,power_law_1.01,0.5489791870117188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,1,power_law_1.01,12.717497253417969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,1,power_law_1.2,0.540780782699585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,1,power_law_1.01,0.562175989151001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,1,power_law_1.2,0.5507711887359619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,1,power_law_1.01,0.5775936126708985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,1,power_law_1.2,0.5836351871490478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,1,power_law_1.01,0.5913728237152099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,1,power_law_1.2,0.6006207942962647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,1,power_law_1.01,0.6215424060821533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,1,power_law_1.01,0.6665152072906494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,1,power_law_1.2,0.6324480056762696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.01,0.7040895938873291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,1,power_law_1.2,0.6374911785125732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,1,power_law_1.2,0.6626239776611328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.01,0.8213760375976562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,1,power_law_1.2,0.7183104038238526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.01,0.9212224006652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,1,power_law_1.2,0.785913610458374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.01,1.1356608390808105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,1,power_law_1.2,0.8521535873413086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.01,1.3509696006774903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,1,power_law_1.2,0.9664447784423829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.01,1.7490432739257813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,1,power_law_1.2,1.0415552139282227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,1,power_law_1.01,2.1701120376586913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,1,power_law_1.2,1.2533760070800781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,1,power_law_1.01,3.0030080795288088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,1,power_law_1.2,1.4332544326782226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,1,power_law_1.01,3.829305648803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,1,power_law_1.2,1.8128639221191407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,1,power_law_1.01,4.645004653930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,1,power_law_1.2,2.2130304336547852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,1,power_law_1.01,7.114598083496094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,1,power_law_1.2,2.9928064346313477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,1,power_law_1.2,3.7586368560791015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,1,power_law_1.01,13.720268249511719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,1,power_law_1.2,4.521260833740234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,1,power_law_1.2,6.817030334472657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,1,power_law_1.2,13.047634887695313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,1,power_law_1.2,0.06914560198783874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,1,power_law_1.2,0.10333440303802491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,1,power_law_1.2,0.08508800268173218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,1,power_law_1.2,0.11794559955596924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,8,balanced,0.06558933357397716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,1,power_law_1.2,0.0983295977115631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,8,balanced,0.06462400158246358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,1,power_law_1.2,0.13165440559387206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,8,balanced,0.09962133566538493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,1,power_law_1.2,0.13095680475234986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,8,balanced,0.06266666452089946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,8,balanced,0.10073066751162212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,8,balanced,0.06480533381303151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,1,power_law_1.2,0.17455359697341918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,8,balanced,0.10124799609184265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,1,power_law_1.2,0.17627520561218263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,8,balanced,0.06972266733646393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,8,balanced,0.10454400380452473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,1,power_law_1.2,0.23705599308013917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,8,balanced,0.06532266736030579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,1,power_law_1.2,0.23029119968414308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,8,balanced,0.10280533631642659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,8,balanced,0.06577066580454509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,1,power_law_1.2,0.29431679248809817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,8,balanced,0.09924800197283427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,1,power_law_1.2,0.26703360080718996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,8,balanced,0.06609599788983662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,8,balanced,0.10175999999046326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,1,power_law_1.2,0.3416192054748535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,8,balanced,0.06922133266925812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,1,power_law_1.2,0.3525183916091919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,8,balanced,0.0999893347422282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,8,balanced,0.0677706648906072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,1,power_law_1.2,0.4486080169677734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,8,balanced,0.09934932986895244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,1,power_law_1.2,0.3621311902999878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,8,balanced,0.06725333134333293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,8,balanced,0.09822932879130046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,1,power_law_1.2,0.46941437721252444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,8,balanced,0.06549333532651265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,1,power_law_1.2,0.3796351909637451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,8,balanced,0.09703999757766724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,8,balanced,0.06620266536871593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,1,power_law_1.2,0.47583999633789065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,8,balanced,0.09428800145785014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,8,balanced,0.06520533561706543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,1,power_law_1.2,0.39336318969726564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,8,balanced,0.09219200412432353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,8,balanced,0.06481599807739258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,1,power_law_1.2,0.505785608291626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,8,balanced,0.09103999535242717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,1,power_law_1.2,0.4148543834686279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,8,balanced,0.06576533118883769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,1,power_law_1.2,0.5164544105529785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,8,balanced,0.08699199557304382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,8,balanced,0.06945600112279256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,1,power_law_1.2,0.4316864013671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,8,balanced,0.08711466193199158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,1,power_law_1.2,0.5325119972229004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,8,balanced,0.08338133494059245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,1,power_law_1.2,0.44748802185058595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,8,balanced,0.08751466870307922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,8,balanced,0.08355200290679932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,1,power_law_1.2,0.5481408119201661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,8,balanced,0.10947199662526448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,1,power_law_1.2,0.4614399909973145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,8,balanced,0.10300800204277039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,1,power_law_1.2,0.5710976123809814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,8,balanced,0.10521599650382996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,1,power_law_1.2,0.4859007835388184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,8,balanced,0.11246933539708455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,8,balanced,0.12777599692344666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,1,power_law_1.2,0.5917503833770752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,1,power_law_1.2,0.5070464134216308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,8,balanced,0.15103466312090555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,8,balanced,0.13869866728782654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,1,power_law_1.2,0.6174399852752686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,1,power_law_1.2,0.5834752082824707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,1,power_law_1.2,0.6863039970397949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,8,balanced,0.19790933529535928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,8,balanced,0.1751520037651062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,1,power_law_1.2,0.6153215885162353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.2,0.74716157913208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,1,power_law_1.2,0.7407167911529541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,8,balanced,0.2330933411916097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,8,balanced,0.24041066567103067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.2,0.858124828338623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,1,power_law_1.2,0.8244223594665527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,8,balanced,0.3205066720644633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.2,0.9699968338012696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,1,power_law_1.2,1.0279232025146485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,8,balanced,0.3044640024503072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.2,1.1818943977355958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,1,power_law_1.2,1.2230400085449218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,8,balanced,0.4059893290201823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,8,balanced,0.43323198954264325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.2,1.3996352195739745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,1,power_law_1.2,1.6129024505615235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.2,1.796518325805664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,1,power_law_1.2,1.9938623428344726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,1,power_law_1.2,2.2105472564697264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,8,balanced,0.5508533318837484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,8,balanced,0.5650186538696289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,1,power_law_1.2,2.768320083618164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,1,power_law_1.2,3.0440704345703127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,1,power_law_1.2,3.51855354309082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,1,power_law_1.2,3.859852981567383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,1,power_law_1.2,4.29510383605957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,8,balanced,0.6767679850260416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,8,balanced,0.7151093482971191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,1,power_law_1.2,4.685395050048828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,1,power_law_1.2,6.57364501953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,1,power_law_1.2,7.134662628173828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,1,power_law_1.2,12.704019165039062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,8,balanced,1.049887975056966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,1,power_law_1.2,13.698431396484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,8,balanced,0.8706453641255697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,8,balanced,1.3129653135935466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,8,balanced,2.0371999740600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,8,balanced,2.489840030670166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,8,balanced,0.06252799928188324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,8,balanced,0.06308799982070923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,8,balanced,0.063509335120519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,8,balanced,0.0641546646753947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,8,balanced,0.06786133348941803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,8,balanced,0.06717333197593689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,8,balanced,0.06351999938488007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,8,balanced,0.06515199939409892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,8,balanced,0.023455999791622162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,8,balanced,0.06437333424886067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,8,balanced,0.024559999505678814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,8,balanced,0.06491200129191081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,8,balanced,0.024122667809327442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,8,balanced,0.06299200157324474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,8,balanced,0.02272533377011617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,8,balanced,0.06338133414586385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,8,balanced,0.06351999938488007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,8,balanced,0.061520000298817955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,8,balanced,0.05952000121275584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,8,power_law_1.01,0.0973695993423462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,8,balanced,0.059893334905306496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,8,power_law_1.01,0.09946240186691284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,8,balanced,0.059157331784566246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,8,power_law_1.01,0.09738240242004395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,8,balanced,0.0739519993464152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,8,power_law_1.01,0.10099200010299683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,8,balanced,0.0692853331565857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,8,power_law_1.01,0.1007807970046997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,8,power_law_1.01,0.09494400024414062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,8,balanced,0.08627200126647949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,8,power_law_1.01,0.09093120098114013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,8,balanced,0.095551997423172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,8,power_law_1.01,0.09010559916496277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,8,power_law_1.01,0.08780159950256347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,8,balanced,0.12237333257993062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,8,power_law_1.01,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,8,power_law_1.01,0.08039680123329163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,8,balanced,0.14485866824785867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,8,power_law_1.01,0.0915391981601715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,8,power_law_1.01,0.10071680545806885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,8,balanced,0.19164800643920898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,8,power_law_1.01,0.10446720123291016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,8,power_law_1.01,0.1118016004562378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,8,power_law_1.01,0.11999360322952271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,8,balanced,0.23642667134602866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,8,power_law_1.01,0.13982080221176146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,8,power_law_1.01,0.1639680027961731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,8,power_law_1.01,0.22050559520721436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,8,power_law_1.01,0.28882560729980467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,8,power_law_1.01,0.062009602785110474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,8,power_law_1.01,0.3146752119064331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,8,balanced,0.3296053409576416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,8,power_law_1.01,0.06495360136032105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,8,power_law_1.01,0.5265600204467773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,8,power_law_1.01,0.06177279949188232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,8,power_law_1.01,0.6567872047424317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,8,power_law_1.01,0.06652160286903382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,8,power_law_1.01,0.8828864097595215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,8,power_law_1.01,0.06511359810829162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,8,power_law_1.01,1.1786304473876954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,8,balanced,0.4238239924112956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,8,power_law_1.01,0.06449919939041138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,8,power_law_1.01,1.7474687576293946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,8,power_law_1.01,0.06190720200538635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,8,power_law_1.01,2.0583168029785157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,8,power_law_1.01,0.06227200031280518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,8,power_law_1.01,2.7161279678344727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,8,power_law_1.01,0.06323199868202209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,8,power_law_1.01,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,8,power_law_1.01,4.489081573486328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,8,balanced,0.511674682299296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,8,power_law_1.01,0.05848960280418396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,8,power_law_1.01,8.611212921142577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,8,power_law_1.01,0.06783360242843628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,8,power_law_1.01,0.07379840016365051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,8,power_law_1.01,0.07138559818267823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,8,power_law_1.01,0.08323839902877808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,8,power_law_1.01,0.07765120267868042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,8,balanced,0.793557325998942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,8,power_law_1.01,0.09259520173072815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,8,power_law_1.01,0.12545280456542968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,8,power_law_1.01,0.1338495969772339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,8,power_law_1.01,0.18488320112228393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,8,power_law_1.01,0.2125632047653198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,8,power_law_1.01,0.28638079166412356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,8,power_law_1.01,0.4043263912200928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,8,power_law_1.01,0.5624063968658447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,8,power_law_1.01,0.7730879783630371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,8,power_law_1.01,1.0834943771362304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,8,power_law_1.01,1.499398422241211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,8,power_law_1.01,1.7739648818969727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,8,balanced,1.5459520022074382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,8,power_law_1.01,2.593337631225586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,8,power_law_1.01,5.390425491333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,8,power_law_1.01,0.06279680132865906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,8,power_law_1.01,0.05988479852676391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,8,power_law_1.01,0.061638402938842776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,8,power_law_1.01,0.06251519918441772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,8,power_law_1.01,0.06451200246810913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,8,power_law_1.01,0.06323840022087097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,8,power_law_1.01,0.06053119897842407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,8,power_law_1.01,0.05952640175819397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,8,power_law_1.01,0.05822719931602478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,8,power_law_1.01,0.05475199818611145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,8,power_law_1.01,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,8,power_law_1.01,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,8,power_law_1.01,0.06442880034446716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,8,power_law_1.01,0.060153597593307497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,8,power_law_1.01,0.07131519913673401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,8,power_law_1.01,0.07208960056304932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,8,balanced,0.02404266595840454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,8,power_law_1.01,0.08664320111274719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,8,balanced,0.027962667246659596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,8,power_law_1.01,0.10047999620437623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,8,balanced,0.044981335600217186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.01,0.12259839773178101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,8,balanced,0.04537599782148997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.01,0.16343679428100585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,8,balanced,0.04762666424115499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.01,0.18967679738998414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,8,power_law_1.2,0.0976960003376007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,8,power_law_1.01,0.02338559925556183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,8,balanced,0.04518933097521464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.01,0.2727231979370117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,8,power_law_1.2,0.09699199795722961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,8,power_law_1.01,0.023398399353027344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,8,balanced,0.039450667798519135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.01,0.3411967992782593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,8,power_law_1.2,0.09648640155792236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,8,power_law_1.01,0.022995199263095855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,8,balanced,0.062021334966023765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.01,0.4705535888671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,8,power_law_1.2,0.1023360013961792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,8,power_law_1.01,0.022067199647426605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,8,balanced,0.0639519989490509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.01,0.566809606552124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,8,power_law_1.2,0.0970687985420227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,8,power_law_1.01,0.022995199263095855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,8,balanced,0.049679999550183616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,8,power_law_1.2,0.09765759706497193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.01,0.944422435760498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,8,power_law_1.01,0.02650879919528961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,8,balanced,0.06741333504517873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,8,power_law_1.2,0.09533439874649048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,8,power_law_1.01,0.04336639940738678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.01,1.1982144355773925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,8,balanced,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,8,power_law_1.2,0.0889855980873108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,8,power_law_1.01,0.04501760005950928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,8,power_law_1.01,1.5584575653076171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,8,balanced,0.06062399844328562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,8,power_law_1.2,0.0870848000049591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,8,power_law_1.01,0.04696959853172302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,8,power_law_1.01,2.390060806274414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,8,balanced,0.08444799979527791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,8,power_law_1.2,0.08540800213813782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,8,power_law_1.01,0.0436928004026413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,8,balanced,0.09433600306510925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,8,power_law_1.2,0.07825279831886292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,8,power_law_1.01,4.909107208251953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,8,power_law_1.01,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,8,power_law_1.2,0.09224960207939148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,8,power_law_1.01,0.06633599996566772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,8,balanced,0.12966932853062949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,8,power_law_1.2,0.09727360010147094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,8,power_law_1.01,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,8,balanced,0.1633386711279551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,8,power_law_1.2,0.10161279439926148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,8,power_law_1.01,0.04764159917831421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,8,power_law_1.2,0.1258687973022461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,8,power_law_1.01,0.06289920210838318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,8,balanced,0.22829866409301758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,8,power_law_1.2,0.12106239795684814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,8,power_law_1.01,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,8,power_law_1.2,0.1461951971054077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,8,power_law_1.01,0.058457601070404056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,8,balanced,0.28782399495442706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,8,power_law_1.2,0.19006719589233398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,8,power_law_1.01,0.0828224003314972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,8,power_law_1.2,0.22048640251159668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.01,0.09270399808883667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,8,balanced,0.4152959982554118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,8,power_law_1.2,0.300217604637146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.01,0.1298624038696289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,8,power_law_1.2,0.3785919904708862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.01,0.16110719442367555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,8,power_law_1.2,0.5125696182250976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.01,0.22809600830078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,8,balanced,0.5361546675364176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,8,power_law_1.2,0.6742080211639404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.01,0.2871167898178101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,8,power_law_1.2,0.9529664039611816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.01,0.41292800903320315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,8,power_law_1.2,1.1599871635437011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.01,0.5360383987426758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,8,balanced,0.7834133307139078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,8,power_law_1.2,1.7503168106079101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.01,0.7826176166534424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.01,1.0270655632019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,8,power_law_1.2,2.3113983154296873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,8,power_law_1.01,1.2743680000305175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,8,power_law_1.2,2.866752052307129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,8,power_law_1.01,2.0115711212158205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,8,power_law_1.2,4.526150512695312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,8,balanced,1.0297173659006755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,8,power_law_1.01,3.990777587890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,8,power_law_1.2,8.505779266357422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,8,balanced,1.2776213486989338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,8,balanced,2.010863939921061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,8,balanced,3.989962577819824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,8,power_law_1.2,0.06285439729690552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,8,power_law_1.2,0.061843198537826535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,8,power_law_1.2,0.06297600269317627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,8,power_law_1.2,0.06474239826202392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,8,power_law_1.2,0.06605439782142639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,8,power_law_1.2,0.06435840129852295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,8,power_law_1.2,0.06291840076446534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,8,power_law_1.2,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,8,power_law_1.2,0.061612802743911746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,8,power_law_1.2,0.058278399705886844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,8,power_law_1.2,0.05567359924316406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,8,power_law_1.2,0.06910719871520996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,8,power_law_1.2,0.06542720198631287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,8,power_law_1.2,0.07156479954719544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,8,power_law_1.2,0.0852288007736206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,8,power_law_1.2,0.08746240139007569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,8,power_law_1.2,0.09717119932174682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,8,power_law_1.2,0.11969280242919922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,8,power_law_1.2,0.13661439418792726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,8,power_law_1.2,0.1851263999938965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,8,power_law_1.2,0.21841280460357665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,8,power_law_1.2,0.3260607957839966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,8,power_law_1.2,0.3908031940460205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,8,power_law_1.2,0.060819202661514284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,8,power_law_1.2,0.5277120113372803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,8,power_law_1.2,0.06048640012741089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,8,power_law_1.2,0.7625088214874267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,8,power_law_1.2,0.023475199937820435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,8,power_law_1.2,0.06172800064086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,8,power_law_1.2,1.1389311790466308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,8,power_law_1.2,0.02322559952735901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,8,power_law_1.2,0.06508799791336059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,8,power_law_1.2,1.4075072288513184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,8,power_law_1.2,0.02366719990968704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,8,power_law_1.2,0.0640447974205017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,8,power_law_1.2,0.021823999285697938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,8,power_law_1.2,1.8568832397460937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,8,power_law_1.2,0.06234880089759827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,8,power_law_1.2,0.023507200181484222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,8,power_law_1.2,2.9271936416625977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,8,power_law_1.2,0.06178560256958008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,8,power_law_1.2,0.02643199861049652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,8,power_law_1.2,0.060652798414230345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,8,power_law_1.2,5.794585418701172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,8,power_law_1.2,0.044710400700569156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,8,power_law_1.2,0.05716480016708374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,8,power_law_1.2,0.04581120014190674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,8,power_law_1.2,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,8,power_law_1.2,0.047731199860572816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,8,power_law_1.2,0.05237119793891907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,8,power_law_1.2,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,8,power_law_1.2,0.06306560039520263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,8,power_law_1.2,0.04169600009918213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,8,power_law_1.2,0.06405760049819946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,8,power_law_1.2,0.0643392026424408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,8,power_law_1.2,0.060083198547363284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,8,power_law_1.2,0.061382400989532473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,8,power_law_1.2,0.07203199863433837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,8,power_law_1.2,0.047116801142692566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,8,power_law_1.2,0.07215999960899352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,8,power_law_1.2,0.0651968002319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,8,power_law_1.2,0.08471680283546448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,8,power_law_1.2,0.05562880039215088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,8,power_law_1.2,0.0998080015182495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,8,power_law_1.2,0.05803520083427429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.2,0.1226304054260254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,8,power_law_1.2,0.0819711983203888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.2,0.16011519432067872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.2,0.09240959882736206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.2,0.1788607954978943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.2,0.13011200428009034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.2,0.2724607944488525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.2,0.1627071976661682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.2,0.34552319049835206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.2,0.22824320793151856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.2,0.48808960914611815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.2,0.28733439445495607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.2,0.578656005859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.2,0.413375997543335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.2,0.9479552268981933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.2,0.5367296218872071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.2,1.2512255668640138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.2,0.7806784152984619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.2,1.0317440032958984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,8,power_law_1.2,1.5227840423583985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,8,power_law_1.2,1.2745216369628907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,8,power_law_1.2,2.47390079498291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,8,power_law_1.2,2.0087039947509764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,8,power_law_1.2,4.69051513671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,8,power_law_1.2,3.982918548583984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,4,balanced,0.07152000069618225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,4,balanced,0.06783466537793477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,4,balanced,0.0830080012480418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,4,balanced,0.08333866794904073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,4,balanced,0.0782293329636256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,4,balanced,0.08907199899355571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,4,balanced,0.09139200051625569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,4,balanced,0.08216000099976857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,4,balanced,0.08815466364224751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,4,balanced,0.07932800054550171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,4,balanced,0.086517333984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,4,balanced,0.08070933322111766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,4,balanced,0.07771199941635132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,4,balanced,0.07221333185831706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,4,balanced,0.08020266890525818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,4,balanced,0.07550933460394542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,4,balanced,0.07733333110809326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,4,balanced,0.09996267159779866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,4,balanced,0.09468799829483032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,4,balanced,0.11864533027013142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,4,balanced,0.1309599975744883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,4,balanced,0.1734880010286967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,4,balanced,0.20431999365488687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,4,balanced,0.28546667098999023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,4,balanced,0.352725346883138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,4,balanced,0.5104159911473592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,4,balanced,0.6583520174026489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,4,balanced,0.8148213227589926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,4,balanced,0.05092266698678335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,4,balanced,0.04818666477998098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,4,balanced,0.0591786652803421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,4,balanced,0.06378666559855144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,4,balanced,0.05486933390299479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,4,balanced,1.273045301437378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,4,balanced,0.055013333757718406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,4,balanced,0.06321066617965698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,4,balanced,0.055013333757718406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,4,balanced,0.05522133409976959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,4,balanced,0.05649066468079885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,4,balanced,0.05693866809209188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,4,balanced,0.055248002211252846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,4,balanced,0.05327466626962026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,4,balanced,0.054005334774653115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,4,balanced,0.06297599772612254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,4,balanced,0.05909866591294607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,4,balanced,0.061146666606267296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,4,balanced,0.07995733122030894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,4,balanced,2.4954400062561035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,4,balanced,0.08273066580295563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,4,balanced,0.10039466619491577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,4,balanced,0.02184533327817917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,4,balanced,0.04862933357556661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,4,balanced,0.11412800351778667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,4,balanced,0.02362666775782903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,4,balanced,0.05192000170548757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,4,balanced,0.022533332308133442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,4,balanced,0.06459199885527293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,4,balanced,0.14913066228230795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,4,balanced,0.0240639994541804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,4,balanced,0.05986666679382324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,4,balanced,0.02369600037733714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,4,balanced,0.06085333228111267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,4,balanced,0.17855467398961386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,4,balanced,0.023914667467276256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,4,balanced,0.059077332417170204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,4,balanced,0.028213332096735638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,4,balanced,0.02619733413060506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,4,balanced,0.24421334266662598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,4,balanced,0.04488533238569895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,4,balanced,0.04463466505209605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,4,balanced,0.044639999667803444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,4,balanced,0.30188800891240436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,4,balanced,0.04655466477076212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,4,balanced,0.044213334719340004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,4,balanced,0.03712533414363861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,4,balanced,0.047093331813812256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,4,balanced,0.4291946490605672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,4,balanced,0.04489600161711375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,4,balanced,0.046384001771608986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,4,balanced,0.0588266650835673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,4,balanced,0.060991997520128884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,4,balanced,0.0846666693687439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,4,balanced,0.5590720176696777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,4,balanced,0.09968533118565877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,4,balanced,0.1376053293546041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,4,balanced,0.16937599579493204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,4,balanced,0.6908213297526041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,4,balanced,0.24337067206700644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,4,balanced,0.3142346739768982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,4,balanced,0.4528799851735433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,4,balanced,1.0697706540425618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,4,balanced,0.5922240018844604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,4,balanced,0.7356320222218832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,4,balanced,2.093162695566813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,4,balanced,1.1553386847178142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,4,balanced,2.2662132581075034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,4,power_law_1.01,0.07167360186576843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,4,power_law_1.01,0.07361279726028443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,4,power_law_1.01,0.07957760095596314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,4,power_law_1.01,0.08463360071182251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,4,power_law_1.01,0.07970560193061829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,4,power_law_1.01,0.08367999792098998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,4,power_law_1.01,0.08147199749946595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,4,power_law_1.01,0.08151040077209473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,4,power_law_1.01,0.08053119778633118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,4,power_law_1.01,0.07944959998130799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,4,power_law_1.01,0.07834879755973816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,4,power_law_1.01,0.08104959726333619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,4,power_law_1.01,0.08539519906044006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,4,power_law_1.01,0.09271680116653443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,4,power_law_1.01,0.10240639448165893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,4,power_law_1.01,0.10202239751815796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,4,power_law_1.01,0.11542400121688842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,4,power_law_1.01,0.13719040155410767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,4,power_law_1.01,0.1559424042701721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,4,power_law_1.01,0.22211840152740478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,4,power_law_1.01,0.2631360054016113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,4,power_law_1.01,0.35269119739532473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,4,power_law_1.01,0.42483201026916506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,4,power_law_1.01,0.699622392654419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,4,power_law_1.01,0.8207679748535156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,4,power_law_1.01,1.0929792404174805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,4,power_law_1.01,1.624006462097168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,4,power_law_1.01,2.018367958068848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,4,power_law_1.01,3.3185150146484377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,4,power_law_1.01,6.7780609130859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,4,power_law_1.01,0.055430400371551516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,4,power_law_1.01,0.054636800289154054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,4,power_law_1.01,0.05243520140647888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,4,power_law_1.01,0.05248000025749207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,4,power_law_1.01,0.05511040091514587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,4,power_law_1.01,0.05597440004348755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,4,power_law_1.01,0.056531202793121335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,4,power_law_1.01,0.0574783980846405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,4,power_law_1.01,0.05576320290565491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,4,power_law_1.01,0.05807999968528747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,4,power_law_1.01,0.055257600545883176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,4,power_law_1.01,0.06345599889755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,4,power_law_1.01,0.06510720252990723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,4,balanced,0.060975998640060425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,4,power_law_1.01,0.06666240096092224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,4,balanced,0.059477334221204124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,4,power_law_1.01,0.07760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,4,balanced,0.05892266829808553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,4,power_law_1.01,0.08083840012550354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,4,balanced,0.06214400132497152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,4,power_law_1.01,0.0837440013885498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,4,balanced,0.059664001067479454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,4,power_law_1.01,0.09854080080986023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,4,power_law_1.01,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,4,balanced,0.060175999999046326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,4,power_law_1.01,0.12003840208053589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,4,power_law_1.01,0.05267840027809143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,4,balanced,0.06380799909432729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,4,power_law_1.01,0.13844480514526367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,4,power_law_1.01,0.05255039930343628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,4,balanced,0.05415999889373779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,4,power_law_1.01,0.1864127993583679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,4,power_law_1.01,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,4,balanced,0.06379200021425883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,4,power_law_1.01,0.25864319801330565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,4,power_law_1.01,0.05720959901809693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,4,balanced,0.05738133192062378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,4,power_law_1.01,0.27913599014282225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,4,power_law_1.01,0.05795199871063232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,4,balanced,0.05741333464781443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,4,power_law_1.01,0.46520318984985354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,4,power_law_1.01,0.0574400007724762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,4,balanced,0.07357333103815715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,4,power_law_1.01,0.592460823059082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,4,power_law_1.01,0.05623679757118225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,4,power_law_1.01,0.7147264003753662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,4,balanced,0.07157333195209503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,4,power_law_1.01,0.058764797449111936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,4,power_law_1.01,1.0131520271301269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,4,power_law_1.01,0.056415998935699464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,4,balanced,0.0860640009244283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,4,power_law_1.01,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,4,power_law_1.01,1.4434304237365723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,4,balanced,0.1018506685892741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,4,power_law_1.01,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,4,power_law_1.01,2.019756889343262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,4,power_law_1.01,0.06332799792289734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,4,balanced,0.12603200475374857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,4,power_law_1.01,4.227065658569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,4,power_law_1.01,0.06312320232391358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,4,power_law_1.01,0.06646400094032287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,4,balanced,0.15265599886576334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,4,power_law_1.01,0.06988800168037415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,4,power_law_1.01,0.08153600096702576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,4,balanced,0.20164799690246582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,4,power_law_1.01,0.09793279767036438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.01,0.10748159885406494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,4,balanced,0.24979732433954874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.01,0.14391679763793946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.01,0.1679360032081604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.01,0.21841280460357665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.01,0.2516416072845459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,4,balanced,0.3513919909795125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.01,0.4251584053039551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.01,0.46932477951049806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.01,0.6018815994262695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.01,1.0192383766174316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,4,balanced,0.4516853491465251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,4,power_law_1.01,1.2483327865600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,4,power_law_1.01,1.7902656555175782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,4,power_law_1.01,3.8016384124755858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,4,balanced,0.554688016573588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,4,power_law_1.01,0.02072319984436035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,4,power_law_1.01,0.021561600267887115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,4,balanced,0.8523413340250651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,4,power_law_1.01,0.02136960029602051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,4,power_law_1.01,0.021817600727081297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,4,power_law_1.01,0.020921599864959717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,4,power_law_1.01,0.02258560061454773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,4,power_law_1.01,0.027142399549484254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,4,power_law_1.01,0.02492160052061081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,4,power_law_1.01,0.04259200096130371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,4,power_law_1.01,0.04219520092010498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,4,power_law_1.01,0.043289598822593686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,4,power_law_1.01,0.04413439929485321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,4,power_law_1.01,0.042182400822639465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,4,power_law_1.01,0.035795199871063235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,4,balanced,1.663413365681966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,4,power_law_1.01,0.04639999866485596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,4,power_law_1.01,0.04416640102863312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,4,power_law_1.01,0.04570879936218262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,4,power_law_1.01,0.058534401655197146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.01,0.062105602025985716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.01,0.08572800159454345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.01,0.09919360280036926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.01,0.13834240436553955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.01,0.17009279727935792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.01,0.2444607973098755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.01,0.31508479118347166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.01,0.454150390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.01,0.5916672229766846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,4,power_law_1.01,0.7348415851593018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,4,power_law_1.01,1.154150390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,4,power_law_1.01,2.271808052062988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,4,power_law_1.2,0.07781760096549988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,4,power_law_1.2,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,4,power_law_1.2,0.07415040135383606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,4,power_law_1.2,0.0818880021572113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,4,power_law_1.2,0.08128640055656433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,4,power_law_1.2,0.08323839902877808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,4,power_law_1.2,0.0824895977973938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,4,power_law_1.2,0.08067200183868409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,4,power_law_1.2,0.0821120023727417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,4,power_law_1.2,0.0812608003616333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,4,power_law_1.2,0.07665280103683472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,4,power_law_1.2,0.08698880076408386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,4,power_law_1.2,0.08350719809532166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,4,power_law_1.2,0.09035519957542419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,4,power_law_1.2,0.10144640207290649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,4,power_law_1.2,0.05857920050621033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,4,power_law_1.2,0.09581440091133117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,4,power_law_1.2,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,4,power_law_1.2,0.12012799978256225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,4,power_law_1.2,0.05194240212440491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,4,power_law_1.2,0.1353279948234558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,4,power_law_1.2,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,4,power_law_1.2,0.16102399826049804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,4,power_law_1.2,0.054054397344589236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,4,power_law_1.2,0.22224640846252441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,4,power_law_1.2,0.055027198791503903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,4,power_law_1.2,0.24233601093292237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,4,power_law_1.2,0.055353599786758426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,4,power_law_1.2,0.3150144100189209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,4,power_law_1.2,0.056480002403259275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,4,power_law_1.2,0.43944320678710935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,4,power_law_1.2,0.05633280277252197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,4,power_law_1.2,0.7736832141876221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,4,power_law_1.2,0.059334397315979004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,4,power_law_1.2,0.9058624267578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,4,power_law_1.2,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,4,power_law_1.2,0.9734399795532227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,4,power_law_1.2,0.06364799737930298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,4,power_law_1.2,1.9065280914306642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,4,power_law_1.2,0.06548479795455933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,4,power_law_1.2,1.9395263671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,4,power_law_1.2,0.0679040014743805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,4,power_law_1.2,2.9897344589233397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,4,power_law_1.2,0.07528960108757018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,4,power_law_1.2,0.07555840015411378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,4,power_law_1.2,7.903584289550781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,4,power_law_1.2,0.0843775987625122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,4,power_law_1.2,0.10780800580978393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,4,power_law_1.2,0.10820480585098266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,4,power_law_1.2,0.1461184024810791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,4,power_law_1.2,0.18161920309066773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,4,power_law_1.2,0.286080002784729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,4,power_law_1.2,0.3262144088745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,4,power_law_1.2,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,4,power_law_1.2,0.5224575996398926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,4,power_law_1.2,0.048614400625228885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,4,power_law_1.2,0.057030397653579715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,4,power_law_1.2,0.7305215835571289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,4,power_law_1.2,0.05764480233192444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,4,power_law_1.2,0.8886079788208008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,4,power_law_1.2,0.0578495979309082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,4,power_law_1.2,1.253439998626709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,4,power_law_1.2,0.05711359977722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,4,power_law_1.2,1.5645119667053222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,4,power_law_1.2,0.05668479800224304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,4,power_law_1.2,2.4114048004150392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,4,power_law_1.2,0.05635200142860412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,4,power_law_1.2,5.2054401397705075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,4,power_law_1.2,0.057132798433303836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,4,power_law_1.2,0.05614079833030701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,4,power_law_1.2,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,4,power_law_1.2,0.06328319907188415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,4,power_law_1.2,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,4,power_law_1.2,0.061510401964187625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,4,power_law_1.2,0.07187839746475219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,4,power_law_1.2,0.07482879757881164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,4,power_law_1.2,0.07771520018577575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,4,power_law_1.2,0.020576000213623047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,4,power_law_1.2,0.0986624002456665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,4,power_law_1.2,0.02184319943189621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.2,0.10720640420913696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,4,power_law_1.2,0.020326399803161622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.2,0.13352320194244385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,4,power_law_1.2,0.02131199985742569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,4,power_law_1.2,0.020044800639152528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.2,0.16125439405441283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,4,power_law_1.2,0.021984000504016877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.2,0.2503488063812256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,4,power_law_1.2,0.02725760042667389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.2,0.29713919162750246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,4,power_law_1.2,0.025190401077270507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.2,0.42383360862731934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,4,power_law_1.2,0.0431551992893219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.2,0.62325119972229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,4,power_law_1.2,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.2,0.6952127933502197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,4,power_law_1.2,0.04361599981784821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.2,1.0018495559692382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,4,power_law_1.2,0.04439679980278015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,4,power_law_1.2,1.3244799613952636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,4,power_law_1.2,0.04299519956111908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,4,power_law_1.2,1.7828479766845704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,4,power_law_1.2,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,4,power_law_1.2,0.04573439955711365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,4,power_law_1.2,3.9606529235839845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,4,power_law_1.2,0.043782401084899905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,4,power_law_1.2,0.0462336003780365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,4,power_law_1.2,0.05788159966468811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.2,0.061267197132110596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.2,0.08568320274353028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.2,0.10007679462432861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.2,0.1385472059249878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.2,0.16978559494018555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.2,0.24289278984069823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.2,0.3136192083358765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.2,0.4526847839355469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.2,0.59269118309021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,4,power_law_1.2,0.7337728023529053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,4,power_law_1.2,1.1540351867675782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,4,power_law_1.2,2.2712896347045897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,2,balanced,0.0484799991051356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,2,balanced,0.06347199777762096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,2,balanced,0.07406400144100189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,2,balanced,0.06951466699441274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,2,balanced,0.06671466430028279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,2,balanced,0.06964799761772156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,2,balanced,0.07268266876538594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,2,balanced,0.07803733150164287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,2,balanced,0.07396799822648366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,2,balanced,0.07377066711584727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,2,balanced,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,2,balanced,0.07210666437943776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,2,balanced,0.07398933172225952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,2,balanced,0.06712000072002411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,2,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,2,balanced,0.07124799986680348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,2,balanced,0.04500266909599304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,2,balanced,0.07386666536331177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,2,balanced,0.058245331048965454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,2,balanced,0.0739519993464152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,2,balanced,0.052613332867622375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,2,balanced,0.0999786655108134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,2,balanced,0.05143466591835022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,2,balanced,0.09570667147636414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,2,balanced,0.05282666782538096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,2,balanced,0.05677866439024607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,2,balanced,0.11992533008257548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,2,balanced,0.05216533442338308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,2,balanced,0.1330560048421224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,2,balanced,0.0621066689491272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,2,balanced,0.05518933137257894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,2,balanced,0.17423999309539795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,2,balanced,0.056373332937558494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,2,balanced,0.05323199927806854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,2,balanced,0.20858667294184366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,2,balanced,0.05593066910902659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,2,balanced,0.05541333556175232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,2,balanced,0.055488000313440956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,2,balanced,0.2887253363927205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,2,balanced,0.059248000383377075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,2,balanced,0.06154666841030121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,2,balanced,0.3667360146840413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,2,balanced,0.08240533371766408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,2,balanced,0.08530132969220479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,2,balanced,0.10811733206113179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,2,balanced,0.5302666823069254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,2,balanced,0.11957866946856181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,2,balanced,0.16019200285275778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,2,balanced,0.042250668009122215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,2,balanced,0.19025067488352457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,2,balanced,0.046522667010625206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,2,balanced,0.6983520189921061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,2,balanced,0.05489600201447805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,2,balanced,0.05955199897289276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,2,balanced,0.2698240081469218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,2,balanced,0.05946666498978933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,2,balanced,0.05825066566467285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,2,balanced,0.33981335163116455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,2,balanced,0.05902933577696482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,2,balanced,0.05515199899673462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,2,balanced,0.8624160289764404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,2,balanced,0.060090666015942894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,2,balanced,0.057114665706952415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,2,balanced,0.4914880196253459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,2,balanced,0.05442133545875549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,2,balanced,0.060933331648508705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,2,balanced,0.05755733450253805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,2,balanced,0.056602666775385536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,2,balanced,0.05691733459631602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,2,balanced,0.6374026536941528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,2,balanced,0.060496002435684204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,2,balanced,1.370730717976888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,2,balanced,0.060138667623202004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,2,balanced,0.01924266666173935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,2,balanced,0.0765066643555959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,2,balanced,0.02000533292690913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,2,balanced,0.07522133489449818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,2,balanced,0.7809173266092936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,2,balanced,0.020303999384244282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,2,balanced,0.09598933657010396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,2,balanced,0.022517333428064983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,2,balanced,0.022570667167504627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,2,balanced,0.1076800028483073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,2,balanced,0.023925334215164185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,2,balanced,0.024314666787783306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,2,balanced,0.1407360037167867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,2,balanced,0.02611733227968216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,2,balanced,0.026208000878492992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,2,balanced,0.16742400328318277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,2,balanced,1.2201440334320068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,2,balanced,0.0259253333012263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,2,balanced,0.025445332129796345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,2,balanced,0.22950933376948038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,2,balanced,0.04438933233420054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,2,balanced,0.044309332966804504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,2,balanced,2.690213203430176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,2,balanced,0.04600533346335093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,2,balanced,0.2880213260650635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,2,balanced,0.038218667109807335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,2,balanced,0.03640533238649368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,2,balanced,0.038506666819254555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,2,balanced,0.4104959964752197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,2,balanced,0.04836266736189524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,2,balanced,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,2,balanced,0.0637546678384145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,2,balanced,2.3853279749552407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,2,balanced,0.0693280001481374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,2,balanced,0.5246133406956991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,2,balanced,0.0969493289788564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,2,balanced,0.1128000020980835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,2,balanced,0.639024019241333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,2,balanced,0.156960000594457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,2,balanced,0.1973173419634501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,2,balanced,0.9989493687947592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,2,balanced,0.28413333495457965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,2,balanced,0.36739198366800946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,2,balanced,0.4533333381017049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,2,balanced,1.9529013633728027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,2,balanced,0.708575963973999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,2,balanced,1.3817706108093262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,2,power_law_1.01,0.05485439896583557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,2,power_law_1.01,0.0547327995300293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,2,power_law_1.01,0.06218879818916321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,2,power_law_1.01,0.07525119781494141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,2,power_law_1.01,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,2,power_law_1.01,0.07701119780540466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,2,power_law_1.01,0.0749888002872467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,2,power_law_1.01,0.04273920059204102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,2,power_law_1.01,0.07429119944572449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,2,power_law_1.01,0.04826880097389221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,2,power_law_1.01,0.0757695972919464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,2,power_law_1.01,0.04589439928531647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,2,power_law_1.01,0.07431039810180665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,2,power_law_1.01,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,2,power_law_1.01,0.07345920205116271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,2,power_law_1.01,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,2,power_law_1.01,0.07426559925079346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,2,power_law_1.01,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,2,power_law_1.01,0.08149120211601257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,2,power_law_1.01,0.0553600013256073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,2,power_law_1.01,0.08039680123329163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,2,power_law_1.01,0.05483520030975342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,2,power_law_1.01,0.09118080139160156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,2,power_law_1.01,0.056428802013397214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,2,power_law_1.01,0.09755520224571228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,2,power_law_1.01,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,2,power_law_1.01,0.0991424024105072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,2,power_law_1.01,0.05251200199127197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,2,power_law_1.01,0.10670080184936523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,2,power_law_1.01,0.052876800298690796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,2,power_law_1.01,0.12460800409317016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,2,power_law_1.01,0.06104959845542908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,2,power_law_1.01,0.16451200246810913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,2,power_law_1.01,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,2,power_law_1.01,0.19031039476394654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,2,power_law_1.01,0.07119359970092773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,2,power_law_1.01,0.24023039340972902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,2,power_law_1.01,0.0733568012714386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,2,power_law_1.01,0.30091519355773927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,2,power_law_1.01,0.07905279994010925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,2,power_law_1.01,0.43884801864624023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,2,power_law_1.01,0.09796479940414429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,2,power_law_1.01,0.5189695835113526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,2,power_law_1.01,0.10002559423446655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,2,power_law_1.01,0.8030207633972168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,2,power_law_1.01,0.13017599582672118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,2,power_law_1.01,0.8627391815185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,2,power_law_1.01,0.14721280336380005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,2,power_law_1.01,1.3097023963928223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,2,power_law_1.01,0.22505600452423097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,2,power_law_1.01,2.266464042663574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,2,power_law_1.01,0.2360896110534668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,2,power_law_1.01,4.340639877319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,2,power_law_1.01,0.36700799465179446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,2,power_law_1.01,0.04572800099849701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,2,power_law_1.01,0.4426112174987793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,2,power_law_1.01,0.04581120014190674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,2,power_law_1.01,0.6334015846252441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,2,power_law_1.01,0.05554559826850891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,2,power_law_1.01,0.7984576225280762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,2,power_law_1.01,0.0542527973651886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,2,power_law_1.01,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,2,power_law_1.01,0.9966208457946777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,2,power_law_1.01,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,2,power_law_1.01,1.6863744735717774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,2,power_law_1.01,0.05836799740791321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,2,power_law_1.01,3.709977722167969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,2,power_law_1.01,0.05560960173606873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,2,power_law_1.01,0.05801600217819214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,2,power_law_1.01,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,2,power_law_1.01,0.05779839754104614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,2,power_law_1.01,0.05681279897689819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,2,power_law_1.01,0.06121600270271301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,2,power_law_1.01,0.06045439839363098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,2,power_law_1.01,0.06826879978179931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,2,power_law_1.01,0.07253119945526124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,2,power_law_1.01,0.018624000251293182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,2,power_law_1.01,0.07311999797821045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,2,power_law_1.01,0.019468800723552705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,2,power_law_1.01,0.09147520065307617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,2,power_law_1.01,0.01932159960269928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.01,0.09351680278778077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,2,power_law_1.01,0.019948799908161164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.01,0.11427199840545654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,2,power_law_1.01,0.022431999444961548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.01,0.1343168020248413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,2,power_law_1.01,0.022944000363349915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.01,0.17338240146636963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,2,power_law_1.01,0.023315200209617616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,2,power_law_1.01,0.022867199778556824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.01,0.2164031982421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,2,power_law_1.01,0.024825599789619446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.01,0.3175935983657837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,2,power_law_1.01,0.0247296005487442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.01,0.4206399917602539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,2,power_law_1.01,0.02505599856376648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.01,0.5016895771026612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,2,power_law_1.01,0.04345600008964538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.01,0.713369607925415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,2,power_law_1.01,0.04300160109996796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,2,power_law_1.01,1.0065728187561036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,2,power_law_1.01,0.04321280121803284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,2,power_law_1.01,1.5300543785095215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,2,power_law_1.01,0.037964800000190736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,2,power_law_1.01,2.9968448638916017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,2,power_law_1.01,0.034867200255393985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,2,power_law_1.01,0.036268800497055054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,2,power_law_1.01,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.01,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.01,0.0636672019958496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.01,0.069760000705719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.01,0.09578880071640014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.01,0.1126911997795105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.01,0.15618560314178467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.01,0.19875839948654175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.01,0.2820863962173462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.01,0.3692800045013428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,2,power_law_1.01,0.4533184051513672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,2,power_law_1.01,0.705241584777832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,2,power_law_1.01,1.3802047729492188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,2,power_law_1.2,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,2,power_law_1.2,0.06024320125579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,2,power_law_1.2,0.06116480231285095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,2,power_law_1.2,0.07259520292282104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,2,power_law_1.2,0.07760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,2,power_law_1.2,0.07633280158042907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,2,power_law_1.2,0.07347840070724487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,2,power_law_1.2,0.07370240092277527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,2,power_law_1.2,0.04669440090656281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,2,power_law_1.2,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,2,power_law_1.2,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,2,power_law_1.2,0.07242239713668823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,2,power_law_1.2,0.04407039880752563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,2,power_law_1.2,0.07221760153770447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,2,power_law_1.2,0.05470719933509827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,2,power_law_1.2,0.08088319897651672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,2,power_law_1.2,0.05541759729385376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,2,power_law_1.2,0.07946239709854126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,2,power_law_1.2,0.05272960066795349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,2,power_law_1.2,0.07667840123176575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,2,power_law_1.2,0.05434880256652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,2,power_law_1.2,0.09598079919815064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,2,power_law_1.2,0.057631999254226685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,2,power_law_1.2,0.09278079867362976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,2,power_law_1.2,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,2,power_law_1.2,0.054611200094223024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,2,power_law_1.2,0.09991040229797363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,2,power_law_1.2,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,2,power_law_1.2,0.11125760078430176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,2,power_law_1.2,0.05862399935722351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,2,power_law_1.2,0.13256959915161132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,2,power_law_1.2,0.06208639740943909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,2,power_law_1.2,0.16609280109405516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,2,power_law_1.2,0.06110720038414001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,2,power_law_1.2,0.2109760046005249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,2,power_law_1.2,0.07428479790687562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,2,power_law_1.2,0.26352639198303224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,2,power_law_1.2,0.07667840123176575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,2,power_law_1.2,0.3348095893859863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,2,power_law_1.2,0.07768319845199585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,2,power_law_1.2,0.44389119148254397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,2,power_law_1.2,0.09310719966888428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,2,power_law_1.2,0.5407743930816651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,2,power_law_1.2,0.10670080184936523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,2,power_law_1.2,0.8345664024353028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,2,power_law_1.2,0.1250048041343689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,2,power_law_1.2,1.150937557220459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,2,power_law_1.2,0.16282880306243896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,2,power_law_1.2,1.4968511581420898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,2,power_law_1.2,0.1899839997291565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,2,power_law_1.2,0.04444159865379334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,2,power_law_1.2,2.1234624862670897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,2,power_law_1.2,0.26530559062957765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,2,power_law_1.2,0.046086400747299194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,2,power_law_1.2,4.187385559082031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,2,power_law_1.2,0.3538815975189209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,2,power_law_1.2,0.04826880097389221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,2,power_law_1.2,0.4292928218841553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,2,power_law_1.2,0.05516800284385681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,2,power_law_1.2,0.6815936088562011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,2,power_law_1.2,0.05570560097694397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,2,power_law_1.2,0.05772799849510193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,2,power_law_1.2,0.8949952125549316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,2,power_law_1.2,0.05807999968528747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,2,power_law_1.2,1.1285375595092773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,2,power_law_1.2,0.05541759729385376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,2,power_law_1.2,1.682111930847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,2,power_law_1.2,0.056806397438049314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,2,power_law_1.2,3.2974143981933595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,2,power_law_1.2,0.05600000023841858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,2,power_law_1.2,0.057529598474502563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,2,power_law_1.2,0.057606399059295654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,2,power_law_1.2,0.05941759943962097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,2,power_law_1.2,0.05982080101966858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,2,power_law_1.2,0.06986240148544312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,2,power_law_1.2,0.07169280052185059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,2,power_law_1.2,0.0781440019607544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,2,power_law_1.2,0.08725119829177856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,2,power_law_1.2,0.018489600718021394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.2,0.09488639831542969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,2,power_law_1.2,0.01984640061855316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.2,0.12869759798049926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,2,power_law_1.2,0.01926400065422058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.2,0.1345023989677429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,2,power_law_1.2,0.020268799364566804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.2,0.1813248038291931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,2,power_law_1.2,0.02250239998102188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.2,0.24976639747619628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,2,power_law_1.2,0.024031999707221984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.2,0.32339839935302733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,2,power_law_1.2,0.02396800071001053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.2,0.3931839942932129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,2,power_law_1.2,0.024166400730609893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.2,0.6311872005462646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,2,power_law_1.2,0.025894400477409363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.2,0.8090559959411621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,2,power_law_1.2,0.02504960000514984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,2,power_law_1.2,0.9676608085632324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,2,power_law_1.2,0.02593280076980591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,2,power_law_1.2,0.04322560131549835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,2,power_law_1.2,1.6882816314697267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,2,power_law_1.2,0.04310399889945984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,2,power_law_1.2,3.1600576400756837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,2,power_law_1.2,0.04380159974098206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,2,power_law_1.2,0.039552000164985654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,2,power_law_1.2,0.035123199224472046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,2,power_law_1.2,0.03712640106678009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,2,power_law_1.2,0.04758400022983551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.2,0.049344000220298764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.2,0.06391040086746216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.2,0.06910079717636108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.2,0.0960640013217926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.2,0.11159679889678956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.2,0.15644799470901488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.2,0.19756799936294556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.2,0.2844608068466187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.2,0.36830079555511475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,2,power_law_1.2,0.4526656150817871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,2,power_law_1.2,0.7084224224090576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,2,power_law_1.2,1.3781120300292968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,1,balanced,0.04408533374468485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,1,balanced,0.0562720000743866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,1,balanced,0.06969066460927327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,1,balanced,0.06783999999364217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,1,balanced,0.0672159989674886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,1,balanced,0.06910400092601776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,1,balanced,0.07056533296902974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,1,balanced,0.07390399773915608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,1,balanced,0.07235733171304067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,1,balanced,0.040037333965301514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,1,balanced,0.07071466743946075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,1,balanced,0.0476746658484141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,1,balanced,0.051354666550954185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,1,balanced,0.05341866612434387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,1,balanced,0.05426666637261709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,1,balanced,0.05288533369700114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,1,balanced,0.0555626650651296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,1,balanced,0.05328533550103506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,1,balanced,0.054773335655530296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,1,balanced,0.05514133473237356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,1,balanced,0.05555200080076853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,1,balanced,0.0552106648683548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,1,balanced,0.05685866872469584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,1,balanced,0.05657066901524862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,1,balanced,0.060266668597857155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,1,balanced,0.06376533210277557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,1,balanced,0.07099199791749318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,1,balanced,0.08959999680519104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,1,balanced,0.0962559978167216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,1,balanced,0.12608533104260763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,1,balanced,0.14170133074124655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,1,balanced,0.1904533306757609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,1,balanced,0.04197866717974345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,1,balanced,0.0489333321650823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,1,balanced,0.23334399859110513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,1,balanced,0.05735466877619425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,1,balanced,0.05938666562239329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,1,balanced,0.06051200131575266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,1,balanced,0.3338720003763835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,1,balanced,0.06204266846179962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,1,balanced,0.059802666306495667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,1,balanced,0.057616000374158226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,1,balanced,0.05989866455396017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,1,balanced,0.4280800024668376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,1,balanced,0.05960000058015188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,1,balanced,0.05692266424496969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,1,balanced,0.057034666339556374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,1,balanced,0.05955733358860016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,1,balanced,0.6202613512674967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,1,balanced,0.05909333129723867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,1,balanced,0.06274666885534923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,1,balanced,0.06454933186372121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,1,balanced,0.06724266707897186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,1,balanced,0.08746666709582011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,1,balanced,0.8075893719991049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,1,balanced,0.0923413336277008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,1,balanced,0.1179093321164449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,1,balanced,0.019461333751678467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,1,balanced,0.13712533315022787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,1,balanced,0.020319999506076176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,1,balanced,0.02065066620707512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,1,balanced,0.180351992448171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,1,balanced,0.9926986694335938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,1,balanced,0.022111999491850536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,1,balanced,0.022597332795461018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,1,balanced,0.22299200296401978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,1,balanced,0.024288001159826916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,1,balanced,0.024143998821576435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,1,balanced,0.31114667654037476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,1,balanced,0.024122667809327442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,1,balanced,0.024570666253566742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,1,balanced,0.024666666984558105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,1,balanced,1.557381312052409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,1,balanced,0.025536000728607178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,1,balanced,0.3996800184249878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,1,balanced,0.02593066543340683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,1,balanced,0.026575999955336254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,1,balanced,0.028437333802382152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,1,balanced,0.5725653171539307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,1,balanced,0.04465066889921824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,1,balanced,0.04517866671085358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,1,balanced,0.04678933322429657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,1,balanced,0.0405173326532046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,1,balanced,0.7503680388132731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,1,balanced,0.04348800083001455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,1,balanced,0.05484800040721893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,1,balanced,0.06113600234190623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,1,balanced,0.9283413092295328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,1,balanced,3.060154596964518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,1,balanced,0.07959466675917308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,1,balanced,0.08906666437784831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,1,balanced,0.12450133760770161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,1,balanced,1.449168046315511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,1,balanced,0.1513920029004415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,1,balanced,0.21270400285720825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,1,balanced,0.2723520000775655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,1,balanced,2.8508691787719727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,1,balanced,0.3340373436609904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,1,balanced,0.5191946824391683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,1,balanced,1.0041120052337646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,1,balanced,0.07028799752394359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,1,balanced,0.07123200098673503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,1,balanced,0.07098133365313213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,1,balanced,0.0702400008837382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,1,balanced,0.0759200006723404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,1,balanced,0.077674667040507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,1,balanced,0.07918933530648549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,1,balanced,0.10332799951235454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,1,balanced,0.10452799995740254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,1,balanced,0.13396267096201578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,1,balanced,0.14723733067512512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,1,power_law_1.01,0.047679999470710756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,1,balanced,0.19596266746520996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,1,power_law_1.01,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,1,balanced,0.24094400803248087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,1,power_law_1.01,0.05614719986915588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,1,power_law_1.01,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,1,power_law_1.01,0.07061120271682739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,1,balanced,0.3362613519032796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,1,power_law_1.01,0.07090560197830201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,1,power_law_1.01,0.07209600210189819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,1,power_law_1.01,0.07416319847106934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,1,balanced,0.4362399975458781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,1,power_law_1.01,0.07043200135231018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,1,power_law_1.01,0.07220479846000671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,1,power_law_1.01,0.07191680073738098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,1,balanced,0.6243199904759725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,1,power_law_1.01,0.07175040245056152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,1,power_law_1.01,0.07482879757881164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,1,power_law_1.01,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,1,power_law_1.01,0.08386560082435608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,1,power_law_1.01,0.03975679874420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,1,power_law_1.01,0.08992000222206116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,1,balanced,0.8176000118255615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,1,power_law_1.01,0.040217599272727965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,1,power_law_1.01,0.09839360117912292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,1,power_law_1.01,0.0426367998123169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,1,power_law_1.01,0.10476160049438477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,1,power_law_1.01,0.049983999133110045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,1,power_law_1.01,0.11753599643707276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,1,power_law_1.01,0.0520576000213623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,1,power_law_1.01,0.14142719507217408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,1,power_law_1.01,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,1,power_law_1.01,0.1616320013999939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,1,balanced,1.005120038986206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,1,power_law_1.01,0.05303040146827698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,1,power_law_1.01,0.20817921161651612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,1,power_law_1.01,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,1,power_law_1.01,0.2700416088104248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,1,power_law_1.01,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,1,power_law_1.01,0.3675647974014282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,1,power_law_1.01,0.05256959795951843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,1,power_law_1.01,0.45976958274841306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,1,power_law_1.01,0.05489280223846436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,1,power_law_1.01,0.6399231910705566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,1,power_law_1.01,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,1,power_law_1.01,0.8295167922973633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,1,power_law_1.01,0.05674880146980286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,1,power_law_1.01,1.0190784454345703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,1,power_law_1.01,0.05940480232238769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,1,balanced,1.598416010538737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,1,power_law_1.01,1.6109439849853515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,1,power_law_1.01,0.07043840289115906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,1,power_law_1.01,0.072953599691391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,1,power_law_1.01,3.1590143203735352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,1,power_law_1.01,0.07973759770393371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,1,power_law_1.01,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,1,power_law_1.01,0.10143359899520873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,1,power_law_1.01,0.12551039457321167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,1,power_law_1.01,0.14799360036849976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,1,power_law_1.01,0.19283839464187622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,1,power_law_1.01,0.24180479049682618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,1,power_law_1.01,0.3398911952972412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,1,power_law_1.01,0.43361921310424806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,1,power_law_1.01,0.04039680063724518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,1,power_law_1.01,0.6251391887664794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,1,power_law_1.01,0.043270400166511534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,1,balanced,3.1440159479777017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,1,power_law_1.01,0.8086272239685058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,1,power_law_1.01,0.04917120039463043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,1,power_law_1.01,0.9929663658142089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,1,power_law_1.01,0.059673601388931276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,1,power_law_1.01,1.5647744178771972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,1,power_law_1.01,0.058508801460266116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,1,power_law_1.01,0.05774719715118408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,1,power_law_1.01,3.060479927062988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,1,power_law_1.01,0.05796480178833008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,1,power_law_1.01,0.05793280005455017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,1,power_law_1.01,0.05677440166473389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,1,power_law_1.01,0.06111360192298889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,1,power_law_1.01,0.059308797121047974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,1,power_law_1.01,0.05919359922409058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,1,power_law_1.01,0.06000639796257019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,1,power_law_1.01,0.06291199922561645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,1,power_law_1.01,0.07343360185623168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,1,power_law_1.01,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,1,power_law_1.01,0.07884160280227662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,1,power_law_1.01,0.08911359906196595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.01,0.09996799826622009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.01,0.1221951961517334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.01,0.143449604511261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.01,0.18557440042495726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.01,0.22910718917846679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.01,0.32200961112976073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.01,0.407206392288208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.01,0.5853439807891846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.01,0.7541440010070801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.01,0.9198719978332519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,1,power_law_1.01,1.4508607864379883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,1,power_law_1.01,2.845439910888672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,1,power_law_1.01,0.017772799730300902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,1,power_law_1.01,0.018483200669288637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,1,power_law_1.01,0.01913599967956543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,1,power_law_1.01,0.01969279944896698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,1,power_law_1.01,0.020768000185489653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,1,power_law_1.2,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,1,power_law_1.01,0.02250880002975464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,1,power_law_1.2,0.05246719717979431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,1,power_law_1.01,0.02287999987602234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,1,power_law_1.2,0.054048001766204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,1,power_law_1.01,0.023571200668811798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,1,power_law_1.2,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,1,power_law_1.01,0.02309119999408722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,1,power_law_1.2,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,1,power_law_1.01,0.025337600708007814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,1,power_law_1.2,0.07013760209083557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,1,power_law_1.01,0.024639999866485594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,1,power_law_1.2,0.07182719707489013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,1,power_law_1.01,0.025113600492477416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,1,power_law_1.2,0.07191680073738098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,1,power_law_1.2,0.040428799390792844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,1,power_law_1.2,0.07082880139350892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,1,power_law_1.2,0.04188160002231598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,1,power_law_1.2,0.07024000287055969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,1,power_law_1.2,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,1,power_law_1.2,0.07093120217323304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,1,power_law_1.2,0.05324159860610962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,1,power_law_1.2,0.0729856014251709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,1,power_law_1.2,0.05169919729232788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,1,power_law_1.2,0.07256320118904114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,1,power_law_1.2,0.05184000134468079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,1,power_law_1.2,0.07456640005111695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,1,power_law_1.2,0.05512319803237915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,1,power_law_1.2,0.08520320057868958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,1,power_law_1.2,0.05355520248413086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,1,power_law_1.2,0.09146239757537841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,1,power_law_1.2,0.05309439897537231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,1,power_law_1.2,0.09666560292243957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,1,power_law_1.2,0.05307520031929016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,1,power_law_1.2,0.10410239696502685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,1,power_law_1.2,0.05392000079154968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,1,power_law_1.2,0.05550720095634461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,1,power_law_1.2,0.11880320310592651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,1,power_law_1.2,0.05927039980888367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,1,power_law_1.2,0.136735999584198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,1,power_law_1.2,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,1,power_law_1.2,0.16414079666137696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,1,power_law_1.2,0.06803839802742004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,1,power_law_1.2,0.21610240936279296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,1,power_law_1.2,0.07388799786567687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,1,power_law_1.2,0.2585472106933594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,1,power_law_1.2,0.07943040132522583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,1,power_law_1.2,0.3626688003540039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,1,power_law_1.2,0.09100800156593322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,1,power_law_1.2,0.4488383769989014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,1,power_law_1.2,0.10158079862594604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,1,power_law_1.2,0.6416768074035645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,1,power_law_1.2,0.12480640411376953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,1,power_law_1.2,0.8259391784667969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,1,power_law_1.2,0.14823039770126342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,1,power_law_1.2,1.020025634765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,1,power_law_1.2,0.19204479455947876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,1,power_law_1.2,1.593273639678955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,1,power_law_1.2,0.24067199230194092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,1,power_law_1.2,3.1501888275146483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,1,power_law_1.2,0.33994879722595217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,1,power_law_1.2,0.4333631992340088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,1,power_law_1.2,0.628115177154541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,1,power_law_1.2,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,1,power_law_1.2,0.8099776268005371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,1,power_law_1.2,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,1,power_law_1.2,0.9938624382019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,1,power_law_1.2,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,1,power_law_1.2,1.560211181640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,1,power_law_1.2,0.05940480232238769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,1,power_law_1.2,0.05944960117340088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,1,power_law_1.2,3.062944030761719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,1,power_law_1.2,0.05806080102920532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,1,power_law_1.2,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,1,power_law_1.2,0.058374398946762086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,1,power_law_1.2,0.05743359923362732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,1,power_law_1.2,0.058348798751831056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,1,power_law_1.2,0.057811200618743896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,1,power_law_1.2,0.05901439785957337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,1,power_law_1.2,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,1,power_law_1.2,0.06234880089759827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,1,power_law_1.2,0.07066879868507385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,1,power_law_1.2,0.07211520075798035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,1,power_law_1.2,0.07961599826812744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,1,power_law_1.2,0.09025279879570007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.2,0.10255359411239624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.2,0.12471040487289428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.2,0.14191360473632814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.2,0.18440959453582764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.2,0.22873599529266359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.2,0.3185215950012207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.2,0.40503678321838377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.2,0.5776000022888184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.2,0.7491712093353271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.2,0.9187520027160645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,1,power_law_1.2,1.4351936340332032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,1,power_law_1.2,2.8320255279541016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,1,power_law_1.2,0.017740799486637114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,1,power_law_1.2,0.018374399840831758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,1,power_law_1.2,0.018572799861431122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,1,power_law_1.2,0.019993600249290467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,1,power_law_1.2,0.02287999987602234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,1,power_law_1.2,0.022227199375629426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,8,balanced,0.12938132882118225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,1,power_law_1.2,0.022393600642681123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,8,balanced,0.13493333260218301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,1,power_law_1.2,0.02314240038394928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,8,balanced,0.13572266697883606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,1,power_law_1.2,0.02336000055074692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,8,balanced,0.1342080036799113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,1,power_law_1.2,0.023468799889087677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,8,balanced,0.13090667128562927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,1,power_law_1.2,0.023756800591945647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,8,balanced,0.13871467113494873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,1,power_law_1.2,0.024646399915218352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,8,balanced,0.1352000037829081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,1,power_law_1.2,0.02536959946155548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,8,balanced,0.12826666235923767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,1,power_law_1.2,0.026745599508285523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,8,balanced,0.13477333386739096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,1,power_law_1.2,0.04418559968471527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,8,balanced,0.13009066383043924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,1,power_law_1.2,0.04441600143909454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,8,balanced,0.12609066565831503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,1,power_law_1.2,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,8,balanced,0.12518399953842163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,1,power_law_1.2,0.03839359879493713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,8,balanced,0.11823999881744385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.2,0.04192639887332916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,8,balanced,0.11360533038775127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.2,0.05395200252532959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,8,balanced,0.11038933197657268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.2,0.05990399718284607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,8,balanced,0.10645332932472229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,1,power_law_1.01,0.025804799795150758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.2,0.07846400141716003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,8,balanced,0.10942400495211284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,1,power_law_1.01,0.026412799954414368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.2,0.08833919763565064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,8,balanced,0.1546026666959127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,1,power_law_1.01,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.2,0.1250496029853821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,8,balanced,0.1432319978872935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,1,power_law_1.01,0.044940799474716187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.2,0.14998400211334229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,8,balanced,0.17756267388661703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,1,power_law_1.01,0.046291199326515195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.2,0.21223039627075196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,1,power_law_1.01,0.04015359878540039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,8,balanced,0.2362239956855774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.2,0.27219200134277344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.01,0.04254080057144165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.2,0.3338752031326294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,8,balanced,0.3080480098724365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.01,0.05429120063781738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,8,balanced,0.09032000104586284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,1,power_law_1.2,0.5217919826507569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.01,0.059628802537918094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,8,balanced,0.08293333152929942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,1,power_law_1.2,1.00698881149292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,8,balanced,0.3911733229955037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.01,0.07998080253601074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,8,balanced,0.0849120020866394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,8,balanced,0.08822932839393616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.01,0.08929280042648316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,8,balanced,0.09012266993522644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,8,balanced,0.5516746838887533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.01,0.12499840259552002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,8,balanced,0.08806932965914409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.01,0.15047680139541625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,8,balanced,0.09013332923253377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,8,balanced,0.714570681254069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.01,0.21045761108398436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,8,balanced,0.0885759989420573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.01,0.2747391939163208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,8,balanced,0.09096533060073853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.01,0.3354176044464111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,8,balanced,0.08898666501045227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,8,balanced,1.0228053728739421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,1,power_law_1.01,0.519155216217041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,8,balanced,0.08734400073687236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,8,balanced,0.08533333738644917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,1,power_law_1.01,1.0057408332824707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,8,balanced,0.08417066931724548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,8,balanced,0.08308800061543782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,8,balanced,1.318442662556966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,8,balanced,0.08182399968306224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,8,balanced,0.08248533308506012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,8,balanced,0.08739200234413147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,8,balanced,0.11524266997973125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,8,balanced,1.6004907290140789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,8,balanced,0.10827733079592387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,8,balanced,0.13876266280810037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,8,balanced,0.17051732540130615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,8,balanced,0.2212000091870626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,8,balanced,2.4224692980448403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,8,balanced,0.26826133330663043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,8,balanced,0.37139201164245605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,8,balanced,0.4761600097020467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,8,balanced,0.6644266843795776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,8,balanced,4.415994644165039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,8,balanced,0.8619679609934489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,8,balanced,1.0471999645233154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,8,balanced,1.625637372334798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,8,balanced,3.1297121047973633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,8,balanced,0.07508266468842824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,8,balanced,0.0842133363087972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,8,balanced,0.08547733227411906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,8,balanced,0.08121066788832347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,8,balanced,0.08346133430798848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,8,balanced,0.07944533228874207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,8,balanced,0.08284266789754231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,8,balanced,0.07865599791208903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,8,balanced,0.07977066437403361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,8,balanced,0.026704000929991405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,8,balanced,0.08153066535790761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,8,balanced,0.028512001037597656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,8,balanced,0.08167999982833862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,8,balanced,0.026709333062171936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,8,balanced,0.08043733239173889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,8,balanced,0.026378666361172993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,8,balanced,0.0739573339621226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,8,balanced,0.02657066782315572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,8,balanced,0.07699733475844066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,8,balanced,0.03161066770553589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,8,balanced,0.07420800129572551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,8,balanced,0.06076799829800924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,8,balanced,0.07147733370463054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,8,balanced,0.06080000102519989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,8,balanced,0.06929600238800049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,8,balanced,0.06090133388837179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,8,balanced,0.10371733705202739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,8,balanced,0.05499200026194254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,8,balanced,0.04797333478927612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,8,balanced,0.09409067034721375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,8,balanced,0.08666132887204488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,8,balanced,0.10849066575368245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,8,balanced,0.0832426647345225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,8,balanced,0.06266666452089946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,8,balanced,0.1264959971110026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,8,balanced,0.09648533662160237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,8,balanced,0.1646506687005361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,8,balanced,0.07531733314196269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,8,balanced,0.09445333480834961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,8,balanced,0.20134933789571127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,8,balanced,0.11538666486740112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,8,balanced,0.14111999670664468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,8,balanced,0.26730666557947796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,8,balanced,0.1953279972076416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,8,balanced,0.24502933025360107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,8,balanced,0.3392053445180257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,8,balanced,0.3471146821975708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,8,power_law_1.01,0.13735040426254272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,8,balanced,0.44312532742818195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,8,balanced,0.48046934604644775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,8,power_law_1.01,0.13118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,8,power_law_1.01,0.12595200538635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,8,balanced,0.6440426508585612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,8,power_law_1.01,0.13937920331954956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,8,power_law_1.01,0.1308351993560791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,8,power_law_1.01,0.12739839553833007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,8,balanced,0.8351946671803793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,8,balanced,0.6194773515065511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,8,power_law_1.01,0.12242560386657715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,8,power_law_1.01,0.12037119865417481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,8,power_law_1.01,0.1133504033088684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,8,balanced,1.235141356786092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,8,power_law_1.01,0.1072767972946167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,8,power_law_1.01,0.10577280521392822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,8,balanced,0.7618079980214437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,8,power_law_1.01,0.13605120182037353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,8,power_law_1.01,0.13524479866027833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,8,power_law_1.01,0.1421504020690918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,8,balanced,1.6324159304300945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,8,power_law_1.01,0.15538560152053832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,8,power_law_1.01,0.17249280214309692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,8,power_law_1.01,0.22510080337524413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,8,power_law_1.01,0.2865151882171631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,8,power_law_1.01,0.36803200244903567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,8,balanced,1.1934666633605957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,8,balanced,2.0233972867329917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,8,power_law_1.01,0.4811647891998291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,8,power_law_1.01,0.6845888137817383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,8,power_law_1.01,0.9066368103027344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,8,power_law_1.01,1.1698944091796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,8,power_law_1.01,1.6765760421752929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,8,power_law_1.01,2.1709503173828124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,8,balanced,3.1980320612589517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,8,power_law_1.01,3.135136032104492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,8,power_law_1.01,4.1523582458496096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,8,power_law_1.01,0.0802944004535675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,8,power_law_1.01,0.08048639893531799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,8,power_law_1.01,4.778239822387695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,8,power_law_1.01,0.08451840281486511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,8,power_law_1.01,7.95888671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,8,power_law_1.01,0.08490239977836608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,8,power_law_1.01,0.08597760200500489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,8,balanced,2.3479894002278647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,8,power_law_1.01,13.820819091796874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,8,power_law_1.01,0.08432000279426574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,8,power_law_1.01,0.08315520286560059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,8,power_law_1.01,0.08113920092582702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,8,power_law_1.01,0.07929599881172181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,8,power_law_1.01,0.07526400089263915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,8,balanced,6.352538426717122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,8,power_law_1.01,0.07691519856452941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,8,power_law_1.01,0.0821951985359192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,8,power_law_1.01,0.09418879747390747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,8,power_law_1.01,0.09588479995727539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,8,power_law_1.01,0.11684479713439941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,8,power_law_1.01,0.1186303973197937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,8,power_law_1.01,0.1351359963417053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,8,power_law_1.01,0.17067519426345826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,8,power_law_1.01,0.19247360229492189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,8,power_law_1.01,0.28327040672302245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,8,power_law_1.01,0.3429120063781738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,8,power_law_1.01,0.4950719833374023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,8,power_law_1.01,0.6565055847167969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,8,power_law_1.01,0.9384896278381347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,8,power_law_1.01,1.17706241607666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,8,power_law_1.01,1.7841983795166017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,8,power_law_1.01,2.336684799194336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,8,power_law_1.01,2.8139583587646486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,8,power_law_1.01,4.504447937011719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,8,power_law_1.01,8.485401916503907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,8,power_law_1.01,0.026092800498008727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,8,power_law_1.01,0.07278079986572265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,8,power_law_1.2,0.12367360591888428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,8,power_law_1.01,0.027347201108932497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,8,power_law_1.01,0.08092799782752991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,8,power_law_1.2,0.13824000358581542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,8,power_law_1.01,0.02545279860496521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,8,power_law_1.01,0.0828544020652771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,8,power_law_1.2,0.1296447992324829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,8,power_law_1.01,0.02426239997148514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,8,power_law_1.01,0.08103039860725403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,8,power_law_1.2,0.13057279586791992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,8,power_law_1.01,0.02545279860496521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,8,power_law_1.01,0.0821120023727417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,8,power_law_1.01,0.028755199909210206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,8,power_law_1.2,0.13545600175857545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,8,power_law_1.01,0.07949439883232116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,8,power_law_1.01,0.05554559826850891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,8,power_law_1.2,0.1308735966682434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,8,power_law_1.01,0.057036799192428586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,8,power_law_1.2,0.12636159658432006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,8,power_law_1.01,0.061363202333450315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,8,power_law_1.2,0.11812479496002197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,8,power_law_1.01,0.05363839864730835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,8,power_law_1.2,0.1087231993675232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,8,power_law_1.01,0.04487679898738861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,8,power_law_1.2,0.1063040018081665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,8,power_law_1.01,0.08628479838371277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,8,power_law_1.2,0.10206079483032227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,8,power_law_1.01,0.07916160225868225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,8,power_law_1.2,0.11603840589523315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,8,power_law_1.01,0.06041600108146668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,8,power_law_1.2,0.1316799998283386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,8,power_law_1.01,0.09132159948348999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,8,power_law_1.2,0.14061440229415895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,8,power_law_1.01,0.07318400144577027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,8,power_law_1.2,0.18276480436325074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,8,power_law_1.01,0.09288319945335388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,8,power_law_1.2,0.17537280321121215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,8,power_law_1.01,0.11604479551315308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,8,power_law_1.2,0.223801589012146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.01,0.13968000411987305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,8,power_law_1.2,0.2912447929382324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.01,0.1942911982536316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,8,power_law_1.2,0.3686144113540649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.01,0.24539520740509033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,8,power_law_1.2,0.4819968223571777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.01,0.34571518898010256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,8,power_law_1.2,0.63471360206604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.01,0.4406720161437988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,8,power_law_1.2,0.9451071739196777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.01,0.6434944152832032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,8,power_law_1.2,1.2091263771057128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,8,power_law_1.2,0.08856319785118102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.01,0.8362112045288086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,8,power_law_1.2,0.08356480002403259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,8,power_law_1.2,1.7965631484985352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.01,1.237657642364502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,8,power_law_1.2,0.08689919710159302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,8,power_law_1.2,2.2650239944458006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.01,1.6259584426879883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,8,power_law_1.2,0.088755202293396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,8,power_law_1.2,3.2854846954345702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,8,power_law_1.01,2.0156288146972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,8,power_law_1.2,0.08663679957389832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,8,power_law_1.2,4.000985717773437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,8,power_law_1.01,3.1941247940063477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,8,power_law_1.2,0.08465279936790467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,8,power_law_1.2,4.993337631225586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,8,power_law_1.2,0.08199040293693542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,8,power_law_1.01,6.3377025604248045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,8,power_law_1.2,0.08212479948997498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,8,power_law_1.2,7.838751983642578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,8,power_law_1.2,0.07925760149955749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,8,power_law_1.2,0.07912319898605347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,8,power_law_1.2,15.105119323730468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,8,power_law_1.2,0.07567999958992004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,8,power_law_1.2,0.10590720176696777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,8,power_law_1.2,0.09262719750404358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,8,power_law_1.2,0.089683198928833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,8,power_law_1.2,0.12391040325164795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,8,power_law_1.2,0.1173248052597046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,8,power_law_1.2,0.1364799976348877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,8,power_law_1.2,0.18368639945983886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,8,power_law_1.2,0.21493759155273437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,8,power_law_1.2,0.28780159950256345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,8,power_law_1.2,0.35311360359191896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,8,power_law_1.2,0.5004032135009766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,8,power_law_1.2,0.6310080051422119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,8,power_law_1.2,0.9358016014099121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,8,power_law_1.2,1.1744256019592285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,8,power_law_1.2,1.737241554260254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,8,power_law_1.2,2.2194623947143555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,8,power_law_1.2,2.8637184143066405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,8,power_law_1.2,4.517740631103516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,8,power_law_1.2,8.794425964355469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,8,power_law_1.2,0.08267520070075988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,8,power_law_1.2,0.07921919822692872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,8,power_law_1.2,0.026790401339530943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,8,power_law_1.2,0.07609599828720093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,8,power_law_1.2,0.026118400692939758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,8,power_law_1.2,0.07848320007324219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,8,power_law_1.2,0.025363200902938844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,8,power_law_1.2,0.08109440207481385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,8,power_law_1.2,0.025209599733352663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,8,power_law_1.2,0.07868160009384155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,8,power_law_1.2,0.02462719976902008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,8,power_law_1.2,0.07610880136489868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,8,power_law_1.2,0.030348798632621764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,8,power_law_1.2,0.0761023998260498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,8,power_law_1.2,0.054579198360443115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,8,power_law_1.2,0.0732479989528656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,8,power_law_1.2,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,8,power_law_1.2,0.06791679859161377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,8,power_law_1.2,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,8,power_law_1.2,0.06386560201644897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,8,power_law_1.2,0.055827200412750244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,8,power_law_1.2,0.0921343982219696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,8,power_law_1.2,0.04547199904918671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,8,power_law_1.2,0.08714879751205444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,8,power_law_1.2,0.08456959724426269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,8,power_law_1.2,0.08041599988937378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,8,power_law_1.2,0.08074880242347718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,8,power_law_1.2,0.10281599760055542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,8,power_law_1.2,0.060262399911880496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,8,power_law_1.2,0.09708799719810486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,8,power_law_1.2,0.09008640050888062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,8,power_law_1.2,0.10705280303955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,8,power_law_1.2,0.0726144015789032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,8,power_law_1.2,0.14494719505310058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,8,power_law_1.2,0.09257599711418152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.2,0.16868480443954467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,8,power_law_1.2,0.1147968053817749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.2,0.2180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.2,0.13928960561752318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.2,0.27619199752807616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.2,0.19471999406814575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.2,0.39559040069580076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.2,0.24754559993743896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,8,power_law_1.01,0.0753279983997345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.2,0.4503168106079102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,4,balanced,0.09725866715113322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.2,0.3469311952590942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,8,power_law_1.01,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.2,0.7439167976379395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,4,balanced,0.09803733229637146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.2,0.442080020904541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,8,power_law_1.01,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,4,balanced,0.10524800419807434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.2,0.9814975738525391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.2,0.6433279991149903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,8,power_law_1.01,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,4,balanced,0.11606933673222859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.2,1.3953344345092773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.2,0.8399871826171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,4,balanced,0.11145066221555074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,8,power_law_1.01,0.06681600213050842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.2,1.853932762145996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.2,1.2333375930786132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,4,balanced,0.11880000432332356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,8,power_law_1.01,0.07685120105743408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,8,power_law_1.2,2.392857551574707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,4,balanced,0.1129919985930125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.2,1.6269119262695313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,8,power_law_1.01,0.08921599984169007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,4,balanced,0.10989333192507426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,8,power_law_1.2,3.805644989013672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,8,power_law_1.2,2.0157888412475584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,8,power_law_1.01,0.08900480270385742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,4,balanced,0.12056533495585124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,8,power_law_1.01,0.10344959497451782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,8,power_law_1.2,3.1976383209228514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,8,power_law_1.2,7.4610748291015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,4,balanced,0.11110400160153706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,8,power_law_1.01,0.10112639665603637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,4,balanced,0.10754666725794475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,8,power_law_1.2,6.338240051269532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,8,power_law_1.01,0.11669759750366211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,4,balanced,0.10156800349553426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,8,power_law_1.01,0.13953919410705568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,4,balanced,0.10021332899729411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.01,0.1797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,4,balanced,0.10225066542625427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.01,0.20599040985107422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,4,balanced,0.09578133622805278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,4,balanced,0.09617599844932556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.01,0.27331840991973877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,4,balanced,0.09800533453623454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.01,0.36411519050598146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,4,balanced,0.14874133467674255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.01,0.5180607795715332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.01,0.716761589050293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,4,balanced,0.13338133692741394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.01,0.8934144020080567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,4,balanced,0.16488533218701681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.01,1.4581567764282226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,4,balanced,0.2055520017941793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.01,1.931283187866211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,4,balanced,0.26874132951100665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,8,power_law_1.01,1.9867071151733398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,8,power_law_1.01,3.6881855010986326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,4,balanced,0.3330666621526082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,8,power_law_1.01,7.397984313964844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,4,balanced,0.4522613286972046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,4,balanced,0.5904959837595621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,4,balanced,0.8369867006937662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,4,balanced,1.089786688486735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,4,balanced,1.3404960632324219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,4,balanced,2.1017866134643555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,4,balanced,4.123978614807129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,4,balanced,0.0582826683918635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,4,balanced,0.06057600180308024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,4,balanced,0.05565333366394043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,4,balanced,0.06316266457239787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,4,balanced,0.08247466882069905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,4,balanced,0.0798880010843277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,4,balanced,0.07342933118343353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,4,balanced,0.07369066774845123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,4,balanced,0.07640000184377034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,4,balanced,0.08239999910195668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,4,balanced,0.07156800230344136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,4,balanced,0.07620266576608022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,4,balanced,0.07607999940713246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,4,balanced,0.08355200290679932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,4,balanced,0.08020266890525818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,4,balanced,0.0786240001519521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,4,balanced,0.07832533121109009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,4,balanced,0.08231466511885326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,4,balanced,0.08186133205890656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,4,balanced,0.07442133128643036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,4,balanced,0.07812800010045369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,4,balanced,0.07367999851703644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,4,balanced,0.07172266642252605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,4,balanced,0.07715733349323273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,4,balanced,0.07126933336257935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,4,balanced,0.07483733197053273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,4,balanced,0.06891733407974243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,4,balanced,0.07302933434645335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,4,balanced,0.07332799832026164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,4,balanced,0.0755626658598582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,4,balanced,0.07141333321730296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,4,balanced,0.07724266747633617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,4,balanced,0.023754666248957317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,4,balanced,0.07066666583220164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,4,balanced,0.08059733112653096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,4,balanced,0.024351999163627625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,4,balanced,0.09342933694521587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,4,balanced,0.10883200168609619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,4,balanced,0.02478933334350586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,4,balanced,0.09072533249855042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,4,balanced,0.10609599947929382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,4,balanced,0.024671999116738636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,4,balanced,0.11320533355077107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,4,balanced,0.13078400492668152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,4,balanced,0.026496000587940216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,4,balanced,0.026821332673231762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,4,balanced,0.12956266601880392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,4,balanced,0.15556800365447998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,4,balanced,0.03309866786003113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,4,balanced,0.02868266652027766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,4,balanced,0.20430932442347208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,4,balanced,0.16724266608556113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,4,balanced,0.05338666836420695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,4,balanced,0.05271466573079427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,4,balanced,0.20043732722600302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,4,balanced,0.25085333983103436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,4,balanced,0.05352533360322317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,4,balanced,0.05572799841562907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,4,balanced,0.2801813284556071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,4,balanced,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,4,balanced,0.3478826681772868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,4,balanced,0.04241600135962168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,4,balanced,0.06632000207901001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,4,balanced,0.35388267040252686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,4,balanced,0.05297600229581197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,4,balanced,0.44832531611124676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,4,balanced,0.06042666733264923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,4,balanced,0.07336000104745229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,4,balanced,0.4965653419494629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,4,balanced,0.0900320013364156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,4,balanced,0.6450613339742025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,4,balanced,0.12222933769226074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,4,balanced,0.1477226714293162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,4,balanced,0.6454240083694458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,4,balanced,0.20569066206614176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,4,balanced,0.8373226324717203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,4,balanced,0.25778132677078247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,4,balanced,0.7915573120117188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,4,balanced,0.36614398161570233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,4,balanced,1.0278879801432292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,4,balanced,0.47313066323598224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,4,balanced,1.238597313563029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,4,balanced,0.6922986507415771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,4,power_law_1.01,0.1086016058921814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,4,power_law_1.01,0.10771839618682862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,4,balanced,1.6178240776062012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,4,power_law_1.01,0.1068992018699646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,4,power_law_1.01,0.11392639875411988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,4,balanced,0.9123253027598063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,4,power_law_1.01,0.11631360054016113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,4,power_law_1.01,0.1124608039855957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,4,power_law_1.01,0.10926079750061035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,4,power_law_1.01,0.10873600244522094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,4,power_law_1.01,0.10848000049591064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,4,power_law_1.01,0.10449919700622559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,4,balanced,1.1297813256581624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,4,power_law_1.01,0.10959999561309815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,4,balanced,2.4369494120279946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,4,power_law_1.01,0.12618240118026733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,4,power_law_1.01,0.11541119813919068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,4,power_law_1.01,0.12072319984436035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,4,power_law_1.01,0.13490560054779052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,4,power_law_1.01,0.13614720106124878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,4,balanced,1.7769279479980469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,4,power_law_1.01,0.1545408010482788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,4,balanced,3.163567860921224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,4,power_law_1.01,0.22094719409942626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,4,power_law_1.01,0.2452608108520508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,4,power_law_1.01,0.3612864017486572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,4,power_law_1.01,0.4052544116973877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,4,power_law_1.01,0.5815040111541748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,4,power_law_1.01,0.6468095779418945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,4,power_law_1.01,1.0251711845397948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,4,power_law_1.01,1.1884480476379395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,4,power_law_1.01,2.0776447296142577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,4,power_law_1.01,2.7979135513305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,4,balanced,3.5155518849690757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,4,power_law_1.01,3.171366310119629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,4,power_law_1.01,5.034195327758789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,4,power_law_1.01,9.043212890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,4,power_law_1.01,0.07528960108757018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,4,power_law_1.01,0.06709120273590088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,4,power_law_1.01,0.07356799840927124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,4,power_law_1.01,0.07521920204162598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,4,power_law_1.01,0.07494400143623352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,4,power_law_1.01,0.07457919716835022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,4,power_law_1.01,0.07583360075950622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,4,power_law_1.01,0.07675520181655884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,4,power_law_1.01,0.06287999749183655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,4,power_law_1.01,0.07888000011444092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,4,power_law_1.01,0.07160320281982421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,4,power_law_1.01,0.07557759881019592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,4,power_law_1.01,0.05852800011634827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,4,power_law_1.01,0.07496320009231568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,4,power_law_1.01,0.0744383990764618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,4,power_law_1.01,0.07642239928245545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,4,power_law_1.01,0.07585920095443725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,4,power_law_1.01,0.08324480056762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,4,power_law_1.01,0.07383040189743043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,4,power_law_1.01,0.08376960158348083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,4,power_law_1.01,0.07560319900512695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,4,power_law_1.01,0.1044927954673767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,4,power_law_1.01,0.07123200297355652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,4,power_law_1.01,0.09560319781303406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,4,power_law_1.01,0.07140480279922486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,4,power_law_1.01,0.11070719957351685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,4,power_law_1.01,0.07370880246162415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,4,power_law_1.01,0.1318719983100891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,4,power_law_1.01,0.07015680074691773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,4,power_law_1.01,0.15383679866790773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,4,power_law_1.01,0.07805439829826355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,4,power_law_1.01,0.19222400188446045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,4,power_law_1.01,0.08190079927444457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,4,power_law_1.01,0.27312641143798827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,4,power_law_1.01,0.07717120051383972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,4,power_law_1.01,0.3605184078216553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,4,power_law_1.01,0.09004160165786743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,4,power_law_1.01,0.487667179107666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,4,power_law_1.2,0.10842880010604858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,4,power_law_1.01,0.09689599871635438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,4,power_law_1.01,0.7428991794586182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,4,power_law_1.2,0.09950720071792603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,4,power_law_1.01,0.10149760246276855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,4,power_law_1.01,0.9119935989379883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,4,power_law_1.01,0.022643199563026427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,4,power_law_1.2,0.10631680488586426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,4,power_law_1.01,0.12221440076828002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,4,power_law_1.01,1.273420810699463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,4,power_law_1.01,0.0240447998046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,4,power_law_1.2,0.10858880281448365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.01,0.1487615942955017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,4,power_law_1.01,1.5930047988891602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,4,power_law_1.01,0.023123200237751006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,4,power_law_1.2,0.11667840480804444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.01,0.19203840494155883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,4,power_law_1.01,2.0376960754394533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,4,power_law_1.01,0.026047998666763307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,4,power_law_1.2,0.10954240560531617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.01,0.22406399250030518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,4,power_law_1.01,0.023686400055885314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,4,power_law_1.01,3.4551231384277346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,4,power_law_1.2,0.11102080345153809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.01,0.26588799953460696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,4,power_law_1.01,0.024902400374412537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,4,power_law_1.2,0.11130239963531494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.01,0.3811455965042114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,4,power_law_1.01,7.379891204833984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,4,power_law_1.01,0.031795200705528257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,4,power_law_1.2,0.10750720500946045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.01,0.5482175827026368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,4,power_law_1.01,0.027475199103355406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,4,power_law_1.2,0.1079103946685791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.01,0.7953152179718017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,4,power_law_1.01,0.052121597528457644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,4,power_law_1.2,0.10111360549926758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.01,1.0524991989135741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,4,power_law_1.01,0.0543615996837616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,4,power_law_1.2,0.121670401096344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.01,1.655731201171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,4,power_law_1.2,0.1284351944923401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,4,power_law_1.01,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,4,power_law_1.01,1.7980735778808594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,4,power_law_1.2,0.12810879945755005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,4,power_law_1.01,0.054028797149658206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,4,power_law_1.2,0.14851839542388917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,4,power_law_1.01,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,4,power_law_1.01,2.657516860961914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,4,power_law_1.2,0.15882879495620728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,4,power_law_1.01,0.04028159976005554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,4,power_law_1.01,6.260409545898438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,4,power_law_1.2,0.16919039487838744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,4,power_law_1.01,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,4,power_law_1.2,0.19267200231552123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,4,power_law_1.01,0.0543615996837616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,4,power_law_1.2,0.24552319049835206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,4,power_law_1.01,0.05973119735717773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,4,power_law_1.2,0.3534080028533936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,4,power_law_1.01,0.07309439778327942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,4,power_law_1.2,0.4446720123291016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.01,0.09038079977035522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,4,power_law_1.2,0.5929791927337646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.01,0.12177920341491699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,4,power_law_1.2,0.8182656288146972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.01,0.14664959907531738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.01,0.20283520221710205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,4,power_law_1.2,1.1005696296691894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.01,0.25547521114349364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,4,power_law_1.2,1.6563711166381836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.01,0.36636159420013426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,4,power_law_1.2,1.965407943725586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.01,0.4758016109466553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,4,power_law_1.2,2.926316833496094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.01,0.6930624008178711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,4,power_law_1.2,3.8373249053955076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.01,0.9112768173217773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,4,power_law_1.2,6.569522857666016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,4,power_law_1.01,1.1239359855651856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,4,power_law_1.01,1.7728895187377929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,4,power_law_1.2,12.847853088378907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,4,power_law_1.01,3.5133502960205076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,4,power_law_1.2,0.06589440107345582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,4,power_law_1.2,0.06568319797515869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,4,power_law_1.2,0.06918399930000305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,4,power_law_1.2,0.07692800164222717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,4,power_law_1.2,0.07616000175476074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,4,power_law_1.2,0.07600640058517456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,4,power_law_1.2,0.0775168001651764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,4,power_law_1.2,0.07459200024604798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,4,power_law_1.2,0.07768319845199585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,4,power_law_1.2,0.0774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,4,power_law_1.2,0.07488639950752259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,4,power_law_1.2,0.07572479844093323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,4,power_law_1.2,0.07399680018424988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,4,power_law_1.2,0.07554559707641602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,4,power_law_1.2,0.07594239711761475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,4,power_law_1.2,0.08462719917297364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,4,power_law_1.2,0.07961599826812744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,4,power_law_1.2,0.08336640000343323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,4,power_law_1.2,0.07691519856452941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,4,power_law_1.2,0.08269439935684204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,4,power_law_1.2,0.07687039971351624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,4,power_law_1.2,0.1074112057685852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,4,power_law_1.2,0.07489920258522034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,4,power_law_1.2,0.09876480102539062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,4,power_law_1.2,0.07312639951705932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,4,power_law_1.2,0.11025279760360718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,4,power_law_1.2,0.07251840233802795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,4,power_law_1.2,0.13995519876480103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,4,power_law_1.2,0.07256960272789001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,4,power_law_1.2,0.17006080150604247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,4,power_law_1.2,0.07934719920158387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,4,power_law_1.2,0.2032639980316162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,4,power_law_1.2,0.08021119832992554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,4,power_law_1.2,0.2969536066055298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,4,power_law_1.2,0.08597760200500489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,4,power_law_1.2,0.021510399878025055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,4,power_law_1.2,0.4364287853240967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,2,balanced,0.06891733407974243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,4,power_law_1.2,0.09144960045814514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,4,power_law_1.2,0.022342400252819063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,4,power_law_1.2,0.5492864131927491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,2,balanced,0.08231466511885326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,4,power_law_1.2,0.096288001537323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,4,power_law_1.2,0.021990400552749634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,4,power_law_1.2,0.7368447780609131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,2,balanced,0.10066133737564087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,4,power_law_1.2,0.10517120361328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,4,power_law_1.2,0.023161600530147552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,4,power_law_1.2,0.8855744361877441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,2,balanced,0.11812266707420349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,4,power_law_1.2,0.022841599583625794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,4,power_law_1.2,0.11710079908370971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,4,power_law_1.2,1.3347328186035157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,2,balanced,0.10318932930628459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,4,power_law_1.2,0.024294400215148927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.2,0.15219199657440186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,2,balanced,0.10258666674296062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,4,power_law_1.2,2.0466239929199217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,4,power_law_1.2,0.030739200115203858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.2,0.1958847999572754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,2,balanced,0.11115200320879619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,4,power_law_1.2,2.1421375274658203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,4,power_law_1.2,0.02791680097579956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.2,0.24416639804840087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,2,balanced,0.10256533821423848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,4,power_law_1.2,4.164787292480469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,4,power_law_1.2,0.051475197076797485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.2,0.2766207933425903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,2,balanced,0.10643200079600017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,4,power_law_1.2,0.05121920108795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.2,0.34771840572357177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,2,balanced,0.10359467069307964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,4,power_law_1.2,8.742137908935547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,4,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,2,balanced,0.10426132877667744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.2,0.52641282081604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,4,power_law_1.2,0.05363839864730835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,2,balanced,0.10240532954533894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.2,0.8276736259460449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,4,power_law_1.2,0.05132160186767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,2,balanced,0.09822932879130046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.2,1.2262656211853027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,4,power_law_1.2,0.04079360067844391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,2,balanced,0.09315733114878337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.2,1.7449024200439454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,4,power_law_1.2,0.06204800009727478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,2,balanced,0.09595732887585957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,4,power_law_1.2,1.9872831344604491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,4,power_law_1.2,0.05244799852371216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,2,balanced,0.09941333532333374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,4,power_law_1.2,3.232838439941406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,4,power_law_1.2,0.05936639904975891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,2,balanced,0.10062399506568909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,4,power_law_1.2,0.07221119999885559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,2,balanced,0.13550399740537009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,4,power_law_1.2,6.9133056640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.2,0.08982399702072144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,2,balanced,0.13079999883969626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.2,0.12202240228652954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,2,balanced,0.16639467080434164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.2,0.14666880369186402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.2,0.20268800258636474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,2,balanced,0.20227199792861938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.2,0.25759360790252683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,2,balanced,0.2608480056126912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.2,0.3670592069625854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.2,0.4735936164855957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,2,balanced,0.31709333260854083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.2,0.6951104164123535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.2,0.9127296447753906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,2,balanced,0.4565066496531169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,4,power_law_1.2,1.1270719528198243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,4,power_law_1.2,1.7799039840698243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,2,balanced,0.5807786782582601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,4,power_law_1.2,3.511199951171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,2,balanced,0.8355680306752523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,2,balanced,1.0989279747009277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,2,balanced,1.3644426663716633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,2,balanced,2.159717400868734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,2,balanced,4.255663871765137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,2,balanced,0.048714667558670044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,2,balanced,0.06105599800745646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,2,balanced,0.06316266457239787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,2,balanced,0.07547733187675476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,2,balanced,0.07188266515731812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,2,balanced,0.07665599882602692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,2,balanced,0.07604266703128815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,2,balanced,0.07619733115037282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,2,balanced,0.07702933251857758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,2,balanced,0.07316266496976216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,2,balanced,0.07318933308124542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,2,balanced,0.0488319993019104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,2,balanced,0.07599466542402904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,2,balanced,0.06109866499900818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,2,balanced,0.06948799888292949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,2,balanced,0.07930666704972585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,2,balanced,0.07178666690985362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,2,balanced,0.07253333429495494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,2,balanced,0.06972266733646393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,2,balanced,0.0787306676308314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,2,balanced,0.07241599758466084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,2,balanced,0.0796319991350174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,2,balanced,0.07189866900444031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,2,balanced,0.08197333415349324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,2,balanced,0.09000533819198608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,2,balanced,0.10193600257237752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,2,balanced,0.07183466851711273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,2,balanced,0.11070400476455688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,2,balanced,0.06991999844710033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,2,balanced,0.0711893339951833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,2,balanced,0.13635733723640442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,2,balanced,0.06749333441257477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,2,balanced,0.16381333271662393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,2,balanced,0.06931733091672261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,2,balanced,0.06724800169467926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,2,balanced,0.21269333362579346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,2,balanced,0.021669333179791767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,2,balanced,0.07233599821726482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,2,balanced,0.022469334304332733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,2,balanced,0.26683733860651654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,2,balanced,0.07134933272997539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,2,balanced,0.023792001108328503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,2,balanced,0.07076799869537354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,2,balanced,0.025413334369659424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,2,balanced,0.37642133235931396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,2,balanced,0.02588266630967458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,2,balanced,0.1027786632378896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,2,balanced,0.027600000301996868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,2,balanced,0.09773866335550944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,2,balanced,0.026677332818508148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,2,balanced,0.4846506516138713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,2,balanced,0.11934933066368103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,2,balanced,0.02826133370399475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,2,balanced,0.02865600089232127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,2,balanced,0.13959999879201254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,2,balanced,0.02828799933195114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,2,balanced,0.6972320079803467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,2,balanced,0.03046400099992752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,2,balanced,0.1823306679725647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,2,balanced,0.05323733389377594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,2,balanced,0.05298133194446564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,2,balanced,0.223471999168396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,2,balanced,0.05401599903901418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,2,balanced,0.9128426710764567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,2,balanced,0.04500266909599304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,2,balanced,0.31278399626413983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,2,balanced,0.04075733323891958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,2,balanced,0.044666667779286705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,2,balanced,0.0574186642964681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,2,balanced,0.3909706672032674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,2,balanced,0.0626453310251236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,2,balanced,1.1353867053985596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,2,balanced,0.081535999973615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,2,balanced,0.10016533732414246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,2,balanced,0.5678079922993978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,2,balanced,0.1330880026022593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,2,balanced,1.7844319343566895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,2,balanced,0.15985600153605142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,2,balanced,0.7364853223164877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,2,balanced,0.22985599438349405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,2,balanced,0.29367999235788983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,2,balanced,0.9094666639963785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,2,balanced,0.42293866475423175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,2,balanced,3.502661387125651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,2,power_law_1.01,0.07546240091323853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,2,power_law_1.01,0.07712000012397766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,2,power_law_1.01,0.08357120156288148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,2,balanced,0.5507733424504598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,2,balanced,1.4156746864318848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,2,power_law_1.01,0.11134719848632812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,2,power_law_1.01,0.10795520544052124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,2,power_law_1.01,0.10420479774475097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,2,power_law_1.01,0.10324480533599853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,2,power_law_1.01,0.10434559583663941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,2,balanced,0.6816266377766927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,2,power_law_1.01,0.10804480314254761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,2,power_law_1.01,0.10232959985733033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,2,power_law_1.01,0.10085120201110839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,2,power_law_1.01,0.10702719688415527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,2,power_law_1.01,0.11720319986343383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,2,power_law_1.01,0.12146559953689576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,2,power_law_1.01,0.12640000581741334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,2,power_law_1.01,0.129203200340271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,2,balanced,2.759530703226725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,2,power_law_1.01,0.13312000036239624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,2,power_law_1.01,0.15474560260772705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,2,balanced,1.064687967300415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,2,power_law_1.01,0.20439679622650148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,2,power_law_1.01,0.2565056085586548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,2,power_law_1.01,0.2815615892410278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,2,power_law_1.01,0.42018561363220214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,2,power_law_1.01,0.5155903816223144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,2,power_law_1.01,0.685868787765503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,2,power_law_1.01,0.8793919563293457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,2,power_law_1.01,1.4813440322875977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,2,power_law_1.01,1.9174463272094726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,2,power_law_1.01,1.9979776382446288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,2,power_law_1.01,3.361568069458008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,2,power_law_1.01,6.596953582763672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,2,balanced,2.092581272125244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,2,power_law_1.01,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,2,power_law_1.01,0.05242239832878113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,2,power_law_1.01,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,2,power_law_1.01,0.07110400199890136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,2,power_law_1.01,0.07363200187683105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,2,power_law_1.01,0.07437440156936645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,2,power_law_1.01,0.07348480224609374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,2,power_law_1.01,0.072953599691391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,2,power_law_1.01,0.07191680073738098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,2,power_law_1.01,0.07345920205116271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,2,power_law_1.01,0.07301120162010193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,2,power_law_1.01,0.07646719813346863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,2,power_law_1.01,0.08097280263900757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,2,power_law_1.01,0.08478080034255982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,2,power_law_1.01,0.09416319727897644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,2,power_law_1.01,0.055199998617172244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,2,power_law_1.01,0.09775999784469605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,2,power_law_1.01,0.05944960117340088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,2,power_law_1.01,0.10431360006332398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,2,power_law_1.01,0.06252800226211548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,2,power_law_1.01,0.12052479982376099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,2,power_law_1.01,0.07468799948692321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,2,power_law_1.01,0.14104959964752198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,2,power_law_1.01,0.070387202501297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,2,power_law_1.01,0.17209600210189818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,2,power_law_1.01,0.07131519913673401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,2,power_law_1.01,0.21765120029449464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,2,power_law_1.01,0.0723583996295929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,2,power_law_1.01,0.298905611038208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,2,power_law_1.01,0.07159039974212647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,2,power_law_1.01,0.37141120433807373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,2,power_law_1.01,0.07132800221443177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,2,power_law_1.01,0.5410751819610595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,2,power_law_1.01,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,2,power_law_1.01,0.6285376071929931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,2,power_law_1.01,0.06993280053138733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,2,power_law_1.01,0.9807552337646485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,2,power_law_1.01,0.07331839799880982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,2,power_law_1.01,0.07907840013504028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,2,power_law_1.01,1.3119104385375977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,2,power_law_1.01,0.07635200023651123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,2,power_law_1.01,1.5808256149291993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,2,power_law_1.01,0.0886784017086029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,2,power_law_1.01,2.725292778015137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,2,power_law_1.01,0.0914687991142273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,2,power_law_1.01,5.1090240478515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,2,power_law_1.01,0.09836159944534302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,2,power_law_1.01,0.019731199741363524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,2,power_law_1.01,0.11178879737854004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,2,power_law_1.01,0.021913599967956544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.01,0.1336832046508789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,2,power_law_1.01,0.02136320024728775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.01,0.16202880144119264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,2,power_law_1.01,0.022387200593948366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.01,0.19080959558486937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,2,power_law_1.01,0.025190401077270507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.01,0.2719360113143921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,2,power_law_1.01,0.02449920028448105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,2,power_law_1.2,0.0767359972000122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.01,0.2987391948699951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,2,power_law_1.01,0.024383999407291412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,2,power_law_1.2,0.0797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.01,0.43091840744018556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,2,power_law_1.01,0.02653439939022064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,2,power_law_1.2,0.08280959725379944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,2,power_law_1.01,0.027820798754692077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.01,0.5180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,2,power_law_1.2,0.10325119495391846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,2,power_law_1.01,0.027270400524139406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.01,0.8273216247558594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,2,power_law_1.2,0.1097983956336975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,2,power_law_1.01,0.029465600848197937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.01,1.1526144027709961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,2,power_law_1.2,0.10738559961318969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,2,power_law_1.01,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,2,power_law_1.01,1.3172479629516602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,2,power_law_1.2,0.10395519733428955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,2,power_law_1.01,0.051577597856521606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,2,power_law_1.2,0.10472320318222046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,2,power_law_1.01,1.9619455337524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,2,power_law_1.01,0.053504002094268796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,2,power_law_1.2,0.10349440574645996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,2,power_law_1.01,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,2,power_law_1.01,3.889459228515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,2,power_law_1.2,0.10565119981765747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,2,power_law_1.01,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,2,power_law_1.2,0.10161919593811035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,2,power_law_1.01,0.04344959855079651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,2,power_law_1.2,0.1081279993057251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,2,power_law_1.01,0.05564799904823303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,2,power_law_1.2,0.10622719526290894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.01,0.06092159748077393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,2,power_law_1.2,0.12039040327072144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.01,0.08127999901771546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,2,power_law_1.2,0.1336832046508789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.01,0.0995199978351593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,2,power_law_1.2,0.13666559457778932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.01,0.13373440504074097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,2,power_law_1.2,0.15158400535583497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.01,0.160697603225708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,2,power_law_1.2,0.16532479524612426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.01,0.23001599311828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,2,power_law_1.2,0.1853119969367981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.01,0.29351680278778075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,2,power_law_1.2,0.24849920272827147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.01,0.42186880111694336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,2,power_law_1.2,0.30499839782714844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.01,0.5489215850830078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,2,power_law_1.2,0.3820672035217285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,2,power_law_1.01,0.6762368202209472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,2,power_law_1.2,0.4965375900268555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,2,power_law_1.01,1.0671168327331544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,2,power_law_1.2,0.7932928085327149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,2,power_law_1.01,2.088492774963379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,2,power_law_1.2,0.9380672454833985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,2,power_law_1.2,1.3611136436462403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,2,power_law_1.2,1.7550207138061524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,2,power_law_1.2,2.2398656845092773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,2,power_law_1.2,3.7030784606933596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,2,power_law_1.2,7.346975708007813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,2,power_law_1.2,0.050432002544403075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,2,power_law_1.2,0.05374720096588135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,2,power_law_1.2,0.05745279788970947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,2,power_law_1.2,0.07190399765968322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,2,power_law_1.2,0.07082880139350892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,2,power_law_1.2,0.0719871997833252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,2,power_law_1.2,0.07249919772148132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,2,power_law_1.2,0.07058560252189636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,2,power_law_1.2,0.07307519912719726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,2,power_law_1.2,0.07479680180549622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,2,power_law_1.2,0.07182719707489013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,2,power_law_1.2,0.07714560031890869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,2,power_law_1.2,0.08001279830932617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,2,power_law_1.2,0.08487679958343505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,2,power_law_1.2,0.09099519848823548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,2,power_law_1.2,0.05777279734611511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,2,power_law_1.2,0.09866880178451538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,2,power_law_1.2,0.056595200300216676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,2,power_law_1.2,0.09818879961967468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,2,power_law_1.2,0.063264000415802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,2,power_law_1.2,0.12595200538635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,2,power_law_1.2,0.07226880192756653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,2,power_law_1.2,0.14584959745407106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,2,power_law_1.2,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,2,power_law_1.2,0.19690879583358764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,2,power_law_1.2,0.0720192015171051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,2,power_law_1.2,0.2129215955734253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,2,power_law_1.2,0.0700543999671936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,2,power_law_1.2,0.28475520610809324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,2,power_law_1.2,0.07055360078811646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,2,power_law_1.2,0.38543999195098877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,2,power_law_1.2,0.07176960110664368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,2,power_law_1.2,0.4736959934234619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,2,power_law_1.2,0.07073280215263367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,2,power_law_1.2,0.6908224105834961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,2,power_law_1.2,0.07340160012245178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,2,power_law_1.2,1.0011455535888671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,2,power_law_1.2,0.07990400195121765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,2,power_law_1.2,1.4050880432128907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,2,power_law_1.2,0.07854080200195312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,2,power_law_1.2,1.6082880020141601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,2,power_law_1.2,0.08243839740753174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,2,power_law_1.2,2.644217681884766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,2,power_law_1.2,0.02008959949016571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,2,power_law_1.2,0.0864512026309967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,2,power_law_1.2,0.02139520049095154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,2,power_law_1.2,0.09448320269584656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,2,power_law_1.2,5.173728179931641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,2,power_law_1.2,0.021478399634361267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,2,power_law_1.2,0.0919871985912323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,2,power_law_1.2,0.023193599283695222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,2,power_law_1.2,0.11649279594421387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,2,power_law_1.2,0.023705600202083586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.2,0.12293119430541992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,2,power_law_1.2,0.02497279942035675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.2,0.14865920543670655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,2,power_law_1.2,0.024928000569343568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.2,0.20193281173706054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,2,power_law_1.2,0.02598400115966797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,1,balanced,0.05959466596444448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.2,0.2507904052734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,2,power_law_1.2,0.02766079902648926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.2,0.32568960189819335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,1,balanced,0.0718399981657664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,2,power_law_1.2,0.02773120105266571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.2,0.409932804107666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,1,balanced,0.09833066662152608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,2,power_law_1.2,0.029203200340270997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.2,0.5797120094299316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,1,balanced,0.10266666611035664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,2,power_law_1.2,0.050220799446105954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.2,0.7505536079406738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,1,balanced,0.09969066580136617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,2,power_law_1.2,0.051648002862930295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.2,1.008556842803955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,1,balanced,0.09479467074076335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,2,power_law_1.2,0.05496320128440857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,1,balanced,0.10240000486373901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,2,power_law_1.2,1.1915264129638672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,2,power_law_1.2,0.04526079893112182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,1,balanced,0.09961066643397014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,2,power_law_1.2,2.262291145324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,2,power_law_1.2,0.040031999349594116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,1,balanced,0.0972213347752889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,2,power_law_1.2,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,2,power_law_1.2,4.4483905792236325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,1,balanced,0.0969599982102712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,2,power_law_1.2,0.05630720257759094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,1,balanced,0.09844266374905904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.2,0.06063359975814819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,1,balanced,0.09490133325258891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.2,0.08014079928398132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,1,balanced,0.09471467137336731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.2,0.10032000541687011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.2,0.13266559839248657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,1,balanced,0.09583999713261922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.2,0.1606528043746948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,1,balanced,0.09876799583435059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.2,0.23001599311828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,1,balanced,0.1016480028629303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.2,0.29271678924560546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,1,balanced,0.10366933544476827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.2,0.4231423854827881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,1,balanced,0.1404213309288025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.2,0.5505663871765136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,1,balanced,0.13869333267211914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,2,power_law_1.2,0.6816319942474365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,1,balanced,0.18071999152501425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,2,power_law_1.2,1.060524845123291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,2,power_law_1.2,2.0903999328613283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,1,balanced,0.21821866432825723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,1,balanced,0.2897226611773173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,1,balanced,0.3530506690343221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,1,balanced,0.5111999909083048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,1,balanced,0.6631733179092407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,1,balanced,0.9522826671600342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,1,balanced,1.2460800011952717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,1,balanced,1.5521386464436848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,1,balanced,2.479215939839681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,1,balanced,0.04407466451327006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,1,balanced,0.05314133564631144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,1,balanced,0.06499733527501424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,1,balanced,0.07565333445866902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,1,balanced,4.92686398824056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,1,balanced,0.07386666536331177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,1,balanced,0.07179733117421468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,1,balanced,0.0705866664648056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,1,balanced,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,1,balanced,0.07223999996980031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,1,balanced,0.0718453327814738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,1,balanced,0.07301333546638489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,1,balanced,0.07457066575686137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,1,balanced,0.07498133182525635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,1,balanced,0.07393600046634674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,1,balanced,0.05220800141493479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,1,balanced,0.05590933561325073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,1,balanced,0.08297599852085114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,1,balanced,0.07145066559314728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,1,balanced,0.08450667063395183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,1,balanced,0.07406400144100189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,1,balanced,0.08867200215657552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,1,balanced,0.06829866766929626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,1,balanced,0.11281599601109822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,1,balanced,0.08269866804281871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,1,balanced,0.07124799986680348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,1,balanced,0.12386133273442586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,1,balanced,0.07340799768765767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,1,balanced,0.15735999743143717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,1,balanced,0.06970666845639546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,1,balanced,0.06795733173688252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,1,balanced,0.18941867351531982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,1,balanced,0.07397333284219106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,1,balanced,0.07054399947325389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,1,balanced,0.25364800294240314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,1,balanced,0.07743999858697255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,1,balanced,0.0732586681842804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,1,balanced,0.3200266758600871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,1,balanced,0.07608533402283986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,1,balanced,0.07749866445859273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,1,balanced,0.45847467581431073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,1,balanced,0.08198399841785431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,1,balanced,0.021210665504137676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,1,balanced,0.10845333337783813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,1,balanced,0.022389332453409832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,1,balanced,0.11161599556605022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,1,balanced,0.5909279982248942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,1,balanced,0.02292266736427943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,1,balanced,0.02274666726589203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,1,balanced,0.13944533467292786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,1,balanced,0.02426133304834366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,1,balanced,0.16659200191497803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,1,balanced,0.02593066543340683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,1,balanced,0.8578346570332845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,1,balanced,0.025802666942278545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,1,balanced,0.2233226696650187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,1,balanced,0.026528000831604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,1,balanced,0.02638400097688039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,1,balanced,0.2752959926923116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,1,balanced,0.026816000541051228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,1,balanced,0.028192001084486645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,1,balanced,1.121066649754842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,1,balanced,0.028917332490285236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,1,balanced,0.39104533195495605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,1,balanced,0.03018666555484136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,1,balanced,0.031146667897701263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,1,balanced,0.05473066866397858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,1,balanced,0.494432012240092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,1,balanced,0.05467733244101206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,1,balanced,0.05667200187842051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,1,balanced,1.3881920178731282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,1,balanced,0.04599999884764353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,1,balanced,0.7068906625111898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,1,balanced,0.04914666712284088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,1,balanced,0.06710933148860931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,1,balanced,0.075013334552447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,1,balanced,0.9440639813741049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,1,balanced,0.10048533479372661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,1,balanced,2.175312042236328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,1,balanced,0.12427199880282085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,1,balanced,0.16932799418767294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,1,balanced,1.1644266446431477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,1,balanced,0.2051253318786621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,1,balanced,0.29847999413808185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,1,balanced,1.8406079610188801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,1,balanced,0.38311998049418133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,1,balanced,4.297967910766602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,1,balanced,0.465557336807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,1,balanced,3.5955254236857095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,1,balanced,0.7274080117543539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,1,power_law_1.01,0.060115200281143186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,1,power_law_1.01,0.06466559767723083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,1,power_law_1.01,0.06762239933013917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,1,power_law_1.01,0.09390720129013061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,1,power_law_1.01,0.09757440090179444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,1,power_law_1.01,0.09901440143585205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,1,power_law_1.01,0.09652479887008666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,1,power_law_1.01,0.09829760193824769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,1,power_law_1.01,0.09522560238838196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,1,power_law_1.01,0.09646080136299133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,1,power_law_1.01,0.09752960205078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,1,power_law_1.01,0.099481600522995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,1,balanced,1.4168586730957031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,1,power_law_1.01,0.102566397190094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,1,power_law_1.01,0.11713279485702514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,1,power_law_1.01,0.12032639980316162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,1,power_law_1.01,0.12957439422607422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,1,power_law_1.01,0.13141119480133057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,1,power_law_1.01,0.14903039932250978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,1,power_law_1.01,0.1765247941017151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,1,power_law_1.01,0.2109760046005249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,1,power_law_1.01,0.23919999599456787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,1,power_law_1.01,0.3211071968078613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,1,power_law_1.01,0.40659198760986326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,1,power_law_1.01,0.5458623886108398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,1,power_law_1.01,0.6890495777130127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,1,power_law_1.01,0.992255973815918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,1,power_law_1.01,1.3093695640563965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,1,power_law_1.01,1.5921024322509765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,1,power_law_1.01,2.485651206970215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,1,power_law_1.01,4.931097412109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,1,power_law_1.01,0.04403199851512909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,1,power_law_1.01,0.04853120148181915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,1,power_law_1.01,0.05424000024795532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,1,power_law_1.01,0.07130879759788514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,1,power_law_1.01,0.06995199918746949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,1,power_law_1.01,0.07150080204010009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,1,power_law_1.01,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,1,power_law_1.01,0.07075200080871583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,1,power_law_1.01,0.07120640277862549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,1,power_law_1.01,0.0710591971874237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,1,power_law_1.01,0.07077119946479797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,1,power_law_1.01,0.07479040026664734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,1,power_law_1.01,0.07697920203208923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,1,power_law_1.01,0.08135679960250855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,1,power_law_1.01,0.09247360229492188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,1,power_law_1.01,0.0932416021823883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,1,power_law_1.01,0.10181119441986083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,1,power_law_1.01,0.11504000425338745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,1,power_law_1.01,0.05120639801025391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,1,power_law_1.01,0.13434879779815673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,1,power_law_1.01,0.05367040038108826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,1,power_law_1.01,0.1676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,1,power_law_1.01,0.05626879930496216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,1,power_law_1.01,0.19566719532012938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,1,power_law_1.01,0.06975359916687011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,1,power_law_1.01,0.2691904067993164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,1,power_law_1.01,0.07218559980392455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,1,power_law_1.01,0.33344640731811526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,1,power_law_1.01,0.07088000178337098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,1,power_law_1.01,0.4696959972381592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,1,power_law_1.01,0.07299200296401978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,1,power_law_1.01,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,1,power_law_1.01,0.6007040023803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,1,power_law_1.01,0.07337599992752075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,1,power_law_1.01,0.8692543983459473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,1,power_law_1.01,0.07329919934272766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,1,power_law_1.01,1.1346495628356934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,1,power_law_1.01,0.07193599939346314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,1,power_law_1.01,1.3958784103393556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,1,power_law_1.01,0.07546240091323853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,1,power_law_1.01,2.185721588134766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,1,power_law_1.01,0.07574399709701538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,1,power_law_1.01,0.01937279999256134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,1,power_law_1.01,4.2888446807861325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,1,power_law_1.01,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,1,power_law_1.01,0.021382400393486024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,1,power_law_1.01,0.08756480216979981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,1,power_law_1.01,0.020857599377632142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,1,power_law_1.01,0.09063040018081665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,1,power_law_1.01,0.021753600239753722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,1,power_law_1.01,0.09794560074806213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,1,power_law_1.01,0.022655999660491942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,1,power_law_1.01,0.1130687952041626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,1,power_law_1.01,0.024160000681877136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.01,0.12492159605026246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,1,power_law_1.01,0.02399359941482544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,1,power_law_1.01,0.025747200846672057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.01,0.15452799797058106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,1,power_law_1.01,0.024857600033283234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.01,0.18021119832992555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,1,power_law_1.01,0.025593599677085875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.01,0.23848960399627686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,1,power_law_1.01,0.026899200677871705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.01,0.29216001033782957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,1,power_law_1.01,0.02757120132446289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.01,0.4001920223236084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,1,power_law_1.01,0.027961599826812743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.01,0.5112959861755371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,1,power_law_1.01,0.02961280047893524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.01,0.7310143947601319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,1,power_law_1.01,0.0525439977645874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.01,0.9654144287109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,1,power_law_1.01,0.05315840244293213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,1,power_law_1.2,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.01,1.1716032028198242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,1,power_law_1.01,0.05581439733505249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,1,power_law_1.2,0.0659775972366333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,1,power_law_1.01,1.8455039978027343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,1,power_law_1.01,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,1,power_law_1.2,0.07428479790687562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.01,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,1,power_law_1.01,3.6048511505126952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,1,power_law_1.2,0.10013439655303955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.01,0.06714879870414733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,1,power_law_1.2,0.09834240078926086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.01,0.07578880190849305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,1,power_law_1.2,0.09756799936294555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.01,0.0992576003074646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,1,power_law_1.2,0.09902719855308532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.01,0.12345600128173828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,1,power_law_1.2,0.09804800152778625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.01,0.17045120000839234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,1,power_law_1.2,0.09811199903488159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.01,0.20646400451660157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,1,power_law_1.2,0.09834880232810975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.01,0.2995392084121704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,1,power_law_1.2,0.09565439820289612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,1,power_law_1.2,0.09619200229644775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.01,0.3824575901031494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,1,power_law_1.2,0.09773439764976502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.01,0.46504960060119627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,1,power_law_1.2,0.11304960250854493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,1,power_law_1.01,0.7256832122802734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,1,power_law_1.2,0.12033920288085938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,1,power_law_1.01,1.4173184394836427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,1,power_law_1.2,0.12259199619293212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,1,power_law_1.2,0.13260159492492676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,1,power_law_1.2,0.14657919406890868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,1,power_law_1.2,0.16892160177230836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,1,power_law_1.2,0.2034048080444336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,1,power_law_1.2,0.23687679767608644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,1,power_law_1.2,0.3219007968902588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,1,power_law_1.2,0.4013184070587158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,1,power_law_1.2,0.532863998413086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,1,power_law_1.2,0.6826623916625977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,1,power_law_1.2,0.9974143981933594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,1,power_law_1.2,1.2785856246948242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,1,power_law_1.2,1.5935680389404296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,1,power_law_1.2,2.4872768402099608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,1,power_law_1.2,4.946739196777344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,1,power_law_1.2,0.04712960124015808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,1,power_law_1.2,0.04790399968624115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,1,power_law_1.2,0.05170559883117676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,1,power_law_1.2,0.06908159852027893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,1,power_law_1.2,0.06967039704322815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,1,power_law_1.2,0.0697983980178833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,1,power_law_1.2,0.07037439942359924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,1,power_law_1.2,0.07008640170097351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,1,power_law_1.2,0.06969599723815918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,1,power_law_1.2,0.07196800112724304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,1,power_law_1.2,0.0688704013824463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,1,power_law_1.2,0.07367039918899536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,1,power_law_1.2,0.07655680179595947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,1,power_law_1.2,0.05121920108795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,1,power_law_1.2,0.07895039916038513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,1,power_law_1.2,0.05384960174560547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,1,power_law_1.2,0.09224960207939148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,1,power_law_1.2,0.05003520250320435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,1,power_law_1.2,0.09432960152626038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,1,power_law_1.2,0.07425280213356018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,1,power_law_1.2,0.09916800260543823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,1,power_law_1.2,0.07425280213356018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,1,power_law_1.2,0.11564799547195434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,1,power_law_1.2,0.0710591971874237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,1,power_law_1.2,0.07082239985466003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,1,power_law_1.2,0.1354367971420288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,1,power_law_1.2,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,1,power_law_1.2,0.17121280431747438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,1,power_law_1.2,0.07190399765968322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,1,power_law_1.2,0.1972864031791687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,1,power_law_1.2,0.0718720018863678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,1,power_law_1.2,0.26641919612884524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,1,power_law_1.2,0.07310720086097718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,1,power_law_1.2,0.3320255994796753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,1,power_law_1.2,0.07560319900512695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,1,power_law_1.2,0.46881918907165526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,1,power_law_1.2,0.07583360075950622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,1,power_law_1.2,0.6017024040222168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,1,power_law_1.2,0.07921919822692872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,1,power_law_1.2,0.8602496147155761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,1,power_law_1.2,0.08751360177993775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,1,power_law_1.2,1.130246353149414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,1,power_law_1.2,0.019628800451755524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,1,power_law_1.2,0.08901119828224183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,1,power_law_1.2,0.021139200031757354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,1,power_law_1.2,1.3894720077514648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,1,power_law_1.2,0.09571840167045594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,1,power_law_1.2,0.02111999988555908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,1,power_law_1.2,2.179180717468262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,1,power_law_1.2,0.10894720554351807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,1,power_law_1.2,0.022694399952888487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,1,power_law_1.2,4.2944385528564455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.2,0.1255231976509094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,1,power_law_1.2,0.02343679964542389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.2,0.1523327946662903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,1,power_law_1.2,0.02494720071554184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.2,0.17802879810333253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,1,power_law_1.2,0.024505600333213806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.2,0.23671679496765136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,1,power_law_1.2,0.026310399174690247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.2,0.2947711944580078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,1,power_law_1.2,0.02515200078487396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.2,0.40113282203674316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,1,power_law_1.2,0.024966399371623992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,1,power_law_1.2,0.02696320116519928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.2,0.5065983772277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,1,power_law_1.2,0.027347201108932497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.2,0.7249343872070313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,1,power_law_1.2,0.027968001365661622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.2,0.9509568214416504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,1,power_law_1.2,0.029139199852943422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.2,1.1755007743835448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,1,power_law_1.2,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,1,power_law_1.2,1.8441919326782226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,1,power_law_1.2,0.053651201725006106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,1,power_law_1.2,3.6017406463623045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,1,power_law_1.2,0.05377280116081238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,1,power_law_1.2,0.04456959962844849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.2,0.04759039878845215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.2,0.06750720143318176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.2,0.07559040188789368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.2,0.09816960096359253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.2,0.12275840044021606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.2,0.16858880519866942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.2,0.20544641017913817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.2,0.29706239700317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.2,0.38252160549163816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.2,0.46334080696105956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,1,power_law_1.2,0.7246272087097168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,1,power_law_1.2,1.4111424446105958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,16,balanced,0.08921600381533305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,16,balanced,0.09091200431187947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,16,balanced,0.10173867146174113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,16,balanced,0.13081600268681845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,16,balanced,0.20101332664489746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,16,balanced,0.3498239914576213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,16,balanced,0.33186666170756024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,16,balanced,0.32311999797821045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,16,balanced,0.33925334612528485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,16,balanced,0.35153067111968994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,16,balanced,0.329312006632487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,16,balanced,0.3252906600634257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,16,balanced,0.3468053340911865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,16,balanced,0.328549325466156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,16,balanced,0.34545600414276123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,16,balanced,0.35066668192545575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,16,balanced,0.3293386697769165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,16,balanced,0.33658134937286377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,16,balanced,0.3370186487833659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,16,balanced,0.3372960090637207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,16,balanced,0.3417760133743286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,16,balanced,0.3593226671218872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,16,balanced,0.3864159981409709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,16,balanced,0.5683306852976481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,16,balanced,0.5634613434473673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,16,balanced,0.06400533517201741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,16,balanced,0.06256533165772755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,16,balanced,0.07205866773923238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,16,balanced,0.09037333726882935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,16,balanced,0.13108799854914346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,16,balanced,0.1960053245226542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,16,balanced,0.19580799341201782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,16,balanced,0.19416000445683798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,16,balanced,0.781440019607544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,16,balanced,0.1933599909146627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,16,balanced,0.19625065724054971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,16,balanced,0.19716266791025797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,16,balanced,0.19161067406336466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,16,balanced,0.1917919913927714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,16,balanced,0.20721600453058878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,16,balanced,0.2083146572113037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,16,balanced,0.2112906575202942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,16,balanced,0.20886399348576865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,16,balanced,0.21820799509684244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,16,balanced,0.06448533137639363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,16,balanced,0.22585066159566244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,16,balanced,0.06325866778691609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,16,balanced,0.07659733295440674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,16,balanced,0.9867200056711832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,16,balanced,0.24580266078313193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,16,balanced,0.1056106686592102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,16,balanced,0.14410666624704996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,16,balanced,0.2336426575978597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,16,balanced,0.256602664788564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,16,balanced,0.21182932456334433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,16,balanced,0.20920000473658243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,16,balanced,0.189626673857371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,16,balanced,0.2964640061060588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,16,balanced,0.20203733444213867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,16,balanced,0.19703465700149536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,16,balanced,0.2013546625773112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,16,balanced,0.194815993309021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,16,balanced,0.33771733442942303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,16,balanced,0.19233600298563638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,16,balanced,0.19790933529535928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,16,balanced,0.21543467044830322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,16,balanced,0.03956266740957896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,16,balanced,0.19851734240849814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,16,balanced,0.03991466760635376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,16,balanced,0.4724800189336141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,16,balanced,0.04533866544564565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,16,balanced,0.21890133619308472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,16,balanced,1.1985066731770833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,16,balanced,0.0483893354733785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,16,balanced,0.060922667384147644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,16,balanced,0.21360532442728677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,16,balanced,0.08401067058245341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,16,balanced,0.097461332877477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,16,balanced,0.09296533465385437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,16,balanced,0.22574400901794434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,16,balanced,0.5336960156758627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,16,balanced,0.13806933164596558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,16,balanced,0.1309866706530253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,16,balanced,0.12155733505884807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,16,balanced,0.22735466559727988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,16,balanced,0.17997332413991293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,16,balanced,0.1826080083847046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,16,balanced,0.18435200055440268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,16,balanced,0.24261333545049033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,16,balanced,0.18585066000620523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,16,balanced,0.18413333098093668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,16,balanced,0.7523946762084961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,16,balanced,0.18020800749460855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,16,balanced,0.20086934169133505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,16,balanced,0.2666986584663391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,16,balanced,0.21710399786631265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,16,balanced,0.25124265750249225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,16,balanced,0.4119573434193929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,16,balanced,0.29282132784525555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,16,balanced,0.9561866919199625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,16,balanced,1.8262453079223633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,16,balanced,0.3640373150507609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,16,balanced,0.4389013449350993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,16,balanced,0.43776532014211017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,16,balanced,0.5861333211263021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,16,balanced,0.6091093222300211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,16,balanced,1.1728160381317139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,16,balanced,0.7444799741109213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,16,balanced,0.7789759635925293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,16,balanced,1.1370773315429688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,16,balanced,1.8068265914916992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,16,balanced,1.4364800453186035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,16,balanced,0.9500426451365153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,16,balanced,3.6124267578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,16,balanced,1.827642599741618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,16,balanced,1.464746634165446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,16,balanced,3.5607999165852866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,16,balanced,2.8080479303995767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,16,balanced,2.8603359858194985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,16,balanced,5.564282735188802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,16,power_law_1.01,0.20194559097290038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,16,power_law_1.01,0.31709439754486085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,16,power_law_1.01,0.1421056032180786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,16,power_law_1.01,0.14945919513702394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,16,power_law_1.01,0.1993407964706421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,16,power_law_1.01,0.22236158847808837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,16,power_law_1.01,0.26959359645843506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,16,power_law_1.01,0.2857408046722412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,16,power_law_1.01,0.3160959959030151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,16,power_law_1.01,0.29291520118713377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,16,power_law_1.01,0.2951103925704956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,16,power_law_1.01,0.314086389541626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,16,power_law_1.01,0.30230400562286375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,16,power_law_1.01,0.29307520389556885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,16,power_law_1.01,0.3186048030853271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,16,power_law_1.01,0.30513920783996584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,16,power_law_1.01,0.13493119478225707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,16,power_law_1.01,0.33138558864593504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,16,power_law_1.01,0.2022847890853882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,16,power_law_1.01,0.328767991065979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,16,power_law_1.01,0.12705279588699342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,16,power_law_1.01,0.368012809753418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,16,power_law_1.01,0.09667199850082397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,16,power_law_1.01,0.4139711856842041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,16,power_law_1.01,0.12291200160980224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,16,power_law_1.01,0.45637121200561526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,16,power_law_1.01,0.14268159866333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,16,power_law_1.01,0.167193603515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,16,power_law_1.01,0.5688960075378418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,16,power_law_1.01,0.1729599952697754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,16,power_law_1.01,0.6220223903656006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,16,power_law_1.01,0.17449599504470825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,16,power_law_1.01,0.8982399940490723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,16,power_law_1.01,0.1698240041732788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,16,power_law_1.01,1.0216896057128906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,16,power_law_1.01,0.18994560241699218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,16,power_law_1.01,1.438380813598633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,16,power_law_1.01,0.18913919925689698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,16,power_law_1.01,1.7609920501708984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,16,power_law_1.01,0.18198399543762206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,16,power_law_1.01,2.2554815292358397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,16,power_law_1.01,0.1804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,16,power_law_1.01,3.4971328735351563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,16,power_law_1.01,0.19503999948501588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,16,power_law_1.01,0.2070080041885376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,16,power_law_1.01,6.6343231201171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,16,power_law_1.01,0.20682880878448487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,16,power_law_1.01,0.22639999389648438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,16,power_law_1.01,0.23899519443511963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,16,power_law_1.01,0.2832576036453247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,16,power_law_1.01,0.3281343936920166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,16,power_law_1.01,0.39140479564666747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,16,power_law_1.01,0.4822400093078613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,16,power_law_1.01,0.6618112087249756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,16,power_law_1.01,0.7925695896148681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,16,power_law_1.01,0.1448192000389099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,16,power_law_1.01,1.0532544136047364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,16,power_law_1.01,0.22510719299316406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,16,power_law_1.01,1.3570048332214355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,16,power_law_1.01,0.10894720554351807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,16,power_law_1.01,1.6548416137695312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,16,power_law_1.01,0.10117119550704956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,16,power_law_1.01,0.1348736047744751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,16,power_law_1.01,2.8265472412109376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,16,power_law_1.01,0.15022720098495485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,16,power_law_1.01,5.13823356628418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,16,power_law_1.01,0.18506239652633666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,16,power_law_1.01,0.17909760475158693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,16,power_law_1.01,0.16758400201797485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,16,power_law_1.01,0.18253439664840698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,16,power_law_1.01,0.17770240306854249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,16,power_law_1.01,0.19296640157699585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,16,power_law_1.01,0.19421440362930298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,16,power_law_1.01,0.046374401450157164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,16,power_law_1.01,0.19178880453109742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,16,power_law_1.01,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,16,power_law_1.01,0.19302400350570678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,16,power_law_1.01,0.046188798546791074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,16,power_law_1.01,0.19956480264663695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,16,power_law_1.01,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,16,power_law_1.01,0.1984063982963562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,16,power_law_1.01,0.060288000106811526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,16,power_law_1.01,0.22093439102172852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,16,power_law_1.01,0.06438400149345398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.01,0.23050880432128906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,16,power_law_1.01,0.07535359859466553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.01,0.27400960922241213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,16,power_law_1.01,0.07503359913825988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,16,power_law_1.01,0.10578559637069702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.01,0.3057471990585327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.01,0.3699712038040161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,16,power_law_1.01,0.10167039632797241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,16,power_law_1.01,0.10806399583816528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.01,0.434003210067749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,16,power_law_1.01,0.13705600500106813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.01,0.5751423835754395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,16,power_law_1.01,0.15304960012435914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.01,0.6734144210815429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,16,power_law_1.01,0.1545215964317322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.01,0.928831958770752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,16,power_law_1.01,0.16424319744110108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.01,1.2613632202148437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,16,power_law_1.01,0.15433599948883056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,16,power_law_1.01,1.4162495613098145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,16,power_law_1.01,0.1697983980178833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,16,power_law_1.01,0.16080000400543212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,16,power_law_1.01,2.34136962890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.01,0.15646079778671265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,16,power_law_1.01,4.684326553344727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.01,0.1739456057548523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.01,0.19059200286865235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.01,0.22403841018676757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.01,0.2676736116409302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.01,0.3740799903869629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.01,0.47844481468200684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.01,0.6459648132324218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.01,0.9227264404296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,16,power_law_1.01,1.1016511917114258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,16,power_law_1.01,1.5957440376281737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,16,power_law_1.01,2.9696767807006834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,16,power_law_1.2,0.19668480157852172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,16,power_law_1.2,0.3306879997253418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,16,power_law_1.2,0.11197439432144166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,16,power_law_1.2,0.1296064019203186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,16,power_law_1.2,0.1762943983078003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,16,power_law_1.2,0.21495680809020995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,16,power_law_1.2,0.27682559490203856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,16,power_law_1.2,0.29360640048980713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,16,power_law_1.2,0.27450239658355713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,16,power_law_1.2,0.12632960081100464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,16,power_law_1.2,0.28457601070404054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,16,power_law_1.2,0.19657599925994873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,16,power_law_1.2,0.3039680004119873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,16,power_law_1.2,0.0765824019908905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,16,power_law_1.2,0.2910207986831665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,16,power_law_1.2,0.09818239808082581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,16,power_law_1.2,0.3177664041519165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,16,power_law_1.2,0.1135103940963745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,16,power_law_1.2,0.1297279953956604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,16,power_law_1.2,0.3193471908569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,16,power_law_1.2,0.17327359914779664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,16,power_law_1.2,0.30468480587005614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,16,power_law_1.2,0.16607999801635742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,16,power_law_1.2,0.3136768102645874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,16,power_law_1.2,0.17448960542678832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,16,power_law_1.2,0.323526406288147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,16,power_law_1.2,0.1779584050178528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,16,power_law_1.2,0.3558207988739014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,16,power_law_1.2,0.17909760475158693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,16,power_law_1.2,0.3646656036376953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,16,power_law_1.2,0.18193279504776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,16,power_law_1.2,0.4444736003875732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,16,power_law_1.2,0.19435520172119142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,16,power_law_1.2,0.5041471958160401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,16,power_law_1.2,0.19941120147705077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,16,power_law_1.2,0.6697855949401855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,16,power_law_1.2,0.19522559642791748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,16,power_law_1.2,0.6936960220336914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,16,power_law_1.2,0.20616319179534912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,16,power_law_1.2,0.9970047950744629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,16,power_law_1.2,0.22056961059570312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,16,power_law_1.2,1.222835159301758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,16,power_law_1.2,0.239136004447937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,16,power_law_1.2,1.6016767501831055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,16,power_law_1.2,0.23407359123229982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,16,power_law_1.2,0.28901760578155516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,16,power_law_1.2,2.225798416137695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,16,power_law_1.2,0.33308799266815187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,16,power_law_1.2,2.73120002746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,16,power_law_1.2,0.44196481704711915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,16,power_law_1.2,4.29884147644043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,16,power_law_1.2,0.5209343910217286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,16,power_law_1.2,0.7177663803100586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,16,power_law_1.2,8.529395294189452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,16,power_law_1.2,0.8810367584228516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,16,power_law_1.2,1.2776639938354493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,16,power_law_1.2,1.6830848693847655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,16,power_law_1.2,1.7967103958129882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,16,power_law_1.2,3.063302421569824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,16,power_law_1.2,5.886016082763672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,16,power_law_1.2,0.1446272015571594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,16,power_law_1.2,0.21561601161956787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,16,power_law_1.2,0.08163840174674988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,16,power_law_1.2,0.1155392050743103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,16,power_law_1.2,0.12599680423736573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,16,power_law_1.2,0.14684159755706788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,16,power_law_1.2,0.15971200466156005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,16,power_law_1.2,0.18247040510177612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,16,power_law_1.2,0.19070719480514525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,16,power_law_1.2,0.18190720081329345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,16,power_law_1.2,0.19059200286865235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,16,power_law_1.2,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,16,power_law_1.2,0.19221760034561158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,16,power_law_1.2,0.18125439882278443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,16,power_law_1.2,0.07203840017318726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,16,power_law_1.2,0.19336960315704346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,16,power_law_1.2,0.04090240001678467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,16,power_law_1.2,0.2020927906036377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,16,power_law_1.2,0.04852479994297028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,16,power_law_1.2,0.2066879987716675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,16,power_law_1.2,0.05115519762039185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,16,power_law_1.2,0.21609599590301515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,16,power_law_1.2,0.056601601839065555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,16,power_law_1.2,0.2345599889755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,16,power_law_1.2,0.07346559762954712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.2,0.26709120273590087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,16,power_law_1.2,0.0782912015914917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,16,power_law_1.2,0.10442880392074586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.2,0.2850303888320923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,16,power_law_1.2,0.10763520002365112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.2,0.3367039918899536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,16,power_law_1.2,0.10185600519180298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.2,0.4057600021362305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,16,power_law_1.2,0.14714879989624025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.2,0.48922882080078123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.2,0.6117887973785401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,16,power_law_1.2,0.14335360527038574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,16,power_law_1.2,0.15327359437942506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.2,0.8610495567321778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,16,power_law_1.2,0.16533119678497316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.2,1.0737088203430176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,16,power_law_1.2,0.16335999965667725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.2,1.2416831970214843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,16,power_law_1.2,0.15857919454574584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,16,power_law_1.2,1.654867172241211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,16,power_law_1.2,0.15010559558868408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,16,power_law_1.2,2.789772796630859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.2,0.15999360084533693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,16,power_law_1.2,5.426822280883789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.2,0.18258559703826904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.2,0.1958400011062622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.2,0.2409343957901001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.2,0.28511359691619875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.2,0.41292800903320315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.2,0.469868803024292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.2,0.7489727973937989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.2,0.9101887702941894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,16,power_law_1.2,1.2482624053955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,16,power_law_1.2,1.8123071670532227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,16,power_law_1.2,3.698233413696289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,8,balanced,0.06638933221499126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,8,balanced,0.07729066908359528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,8,balanced,0.09410666426022847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,8,balanced,0.1229759951432546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,8,balanced,0.1872160037358602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,8,balanced,0.32320000727971393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,8,balanced,0.3179626663525899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,8,balanced,0.05243200063705444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,8,balanced,0.3097493251164754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,8,balanced,0.06116800010204315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,8,balanced,0.31057600180308026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,8,balanced,0.06897599995136261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,8,balanced,0.08755200107892354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,8,balanced,0.125301331281662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,8,balanced,0.20728000005086264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,8,balanced,0.19364267587661743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,8,balanced,0.19483200709025064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,8,balanced,0.18914133310317993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,8,balanced,0.1962719957033793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,8,balanced,0.1845653255780538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,8,balanced,0.18136000633239746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,8,balanced,0.19674134254455566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,8,balanced,0.19026132424672446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,8,balanced,0.19608533382415771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,8,balanced,0.2037973403930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,8,balanced,0.20595733324686685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,8,balanced,0.21924267212549844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,8,balanced,0.22513065735499063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,8,balanced,0.2507999936739604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,8,balanced,0.2680426637331645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,8,balanced,0.3184000054995219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,8,balanced,0.3663146495819092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,8,balanced,0.5089279810587565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,8,balanced,0.5929119984308878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,8,balanced,0.05621333420276642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,8,balanced,0.06741333504517873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,8,balanced,0.0718453327814738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,8,balanced,0.8337013721466064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,8,balanced,0.10045333703358968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,8,balanced,0.14571199814478555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,8,balanced,0.20861866076787314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,8,balanced,0.20644267400105795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,8,balanced,0.19803200165430704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,8,balanced,0.1974560022354126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,8,balanced,0.19001599152882895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,8,balanced,0.18929600715637207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,8,balanced,0.18649599949518839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,8,balanced,0.18768533070882162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,8,balanced,0.19192532698313394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,8,balanced,0.19787200291951498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,8,balanced,1.071552038192749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,8,balanced,0.20982933044433594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,8,balanced,0.19793599843978882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,8,balanced,0.20817599693934122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,8,balanced,0.025562666356563568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,8,balanced,0.027888000011444092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,8,balanced,0.21341866254806519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,8,balanced,0.030741333961486816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,8,balanced,0.04562133550643921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,8,balanced,0.057034666339556374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,8,balanced,0.22382400433222452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,8,balanced,0.08319466809431712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,8,balanced,0.08763733506202698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,8,balanced,0.08893332878748576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,8,balanced,0.24301334222157797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,8,balanced,0.09865599870681763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,8,balanced,0.0974026620388031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,8,balanced,1.314570665359497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,8,balanced,0.10022399822870891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,8,balanced,0.13769066333770752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,8,balanced,0.2634506622950236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,8,balanced,0.13055466612180075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,8,balanced,0.13498666882514954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,8,balanced,0.18476800123850504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,8,balanced,0.18753600120544434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,8,balanced,0.30380799372990924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,8,balanced,0.19803200165430704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,8,balanced,0.20025600989659628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,8,balanced,0.20522665977478027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,8,balanced,0.46209601561228436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,8,balanced,0.23080533742904663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,8,balanced,0.25407467285792035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,8,balanced,0.5049706697463989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,8,balanced,0.29899734258651733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,8,balanced,2.0325600306193032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,8,balanced,0.3450080156326294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,8,balanced,0.7092639605204264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,8,balanced,0.43169601758321124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,8,balanced,0.506986657778422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,8,balanced,0.9184532960255941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,8,balanced,0.3298666675885518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,8,balanced,0.8086187044779459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,8,balanced,0.2961493333180745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,8,balanced,0.3179733355840047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,8,balanced,0.33059199651082355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,8,balanced,0.2951893409093221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,8,balanced,0.33063467343648273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,8,balanced,1.119167963663737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,8,balanced,0.324837327003479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,8,balanced,0.3132159908612569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,8,balanced,0.3247893253962199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,8,balanced,0.31986133257548016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,8,balanced,4.0100053151448565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,8,balanced,0.33513065179189044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,8,balanced,0.9981333414713541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,8,balanced,0.34885334968566895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,8,balanced,0.37934398651123047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,8,balanced,0.40987201531728107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,8,balanced,1.7275627454121907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,8,balanced,0.5936586856842041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,8,balanced,1.2841119766235352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,8,balanced,0.6033386786778768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,8,balanced,0.8532746632893881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,8,balanced,1.9447627067565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,8,balanced,1.0673173268636067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,8,balanced,3.370682716369629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,8,balanced,1.283450682957967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,8,balanced,2.0310773849487305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,8,balanced,3.84931214650472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,8,power_law_1.01,0.11867519617080688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,8,power_law_1.01,0.18492159843444825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,8,power_law_1.01,0.09904000163078308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,8,power_law_1.01,0.13801599740982057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,8,power_law_1.01,0.16197760105133058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,8,power_law_1.01,0.18762880563735962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,8,power_law_1.01,0.24547839164733887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,8,power_law_1.01,0.25043840408325196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,8,power_law_1.01,0.2624959945678711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,8,power_law_1.01,0.2521023988723755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,8,power_law_1.01,0.2661247968673706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,8,power_law_1.01,0.2792896032333374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,8,power_law_1.01,0.2781375885009766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,8,power_law_1.01,0.27456638813018797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,8,power_law_1.01,0.2867072105407715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,8,power_law_1.01,0.2949120044708252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,8,power_law_1.01,0.30427520275115966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,8,power_law_1.01,0.33194239139556886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,8,power_law_1.01,0.3309056043624878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,8,power_law_1.01,0.37792000770568845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,8,power_law_1.01,0.41112961769104006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,8,balanced,3.9755786259969077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,8,power_law_1.01,0.5075263977050781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,8,power_law_1.01,0.6099775791168213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,8,power_law_1.01,0.7429503917694091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,8,power_law_1.01,0.9974080085754394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,8,power_law_1.01,1.317523193359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,8,power_law_1.01,1.590182399749756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,8,power_law_1.01,1.996940803527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,8,power_law_1.01,3.180364799499512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,8,power_law_1.01,5.538835144042968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,8,power_law_1.01,0.08918399810791015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,8,power_law_1.01,0.13320959806442262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,8,power_law_1.01,0.0833791971206665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,8,power_law_1.01,0.08610560297966004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,8,power_law_1.01,0.11383039951324463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,8,power_law_1.01,0.13911039829254152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,8,power_law_1.01,0.15965440273284912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,8,power_law_1.01,0.1685696005821228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,8,power_law_1.01,0.1709887981414795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,8,power_law_1.01,0.16978559494018555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,8,power_law_1.01,0.17984000444412232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,8,power_law_1.01,0.18216960430145263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,8,power_law_1.01,0.18388479948043823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,8,power_law_1.01,0.1868288040161133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,8,power_law_1.01,0.18980480432510377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,8,power_law_1.01,0.1955072045326233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,8,power_law_1.01,0.19159679412841796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,8,power_law_1.01,0.2187391996383667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,8,power_law_1.01,0.24043519496917726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,8,power_law_1.01,0.27983999252319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,8,power_law_1.01,0.3250816106796265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,8,power_law_1.01,0.3982719898223877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,8,power_law_1.01,0.46425600051879884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,8,power_law_1.01,0.6287615776062012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,8,power_law_1.01,0.7771647930145263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,8,power_law_1.01,1.0528960227966309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,8,power_law_1.01,0.10253440141677857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,8,power_law_1.01,1.387667179107666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,8,power_law_1.01,0.1434816002845764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,8,power_law_1.01,1.7460351943969727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,8,power_law_1.01,0.0946943998336792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,8,power_law_1.01,2.588755226135254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,8,power_law_1.01,0.09921919703483581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,8,power_law_1.01,0.1193727970123291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,8,power_law_1.01,4.905324935913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,8,power_law_1.01,0.13494399785995484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,8,power_law_1.01,0.1677183985710144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,8,power_law_1.01,0.17268480062484742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,8,power_law_1.01,0.18026880025863648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,8,power_law_1.01,0.17908480167388915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,8,power_law_1.01,0.1708672046661377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,8,power_law_1.01,0.18471039533615113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,8,power_law_1.01,0.17879040241241456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,8,power_law_1.01,0.18630399703979492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,8,power_law_1.01,0.19738880395889283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,8,power_law_1.01,0.19219839572906494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,8,power_law_1.01,0.20223360061645507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,8,power_law_1.01,0.21400320529937744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.01,0.22892160415649415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.01,0.26666879653930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.01,0.29700479507446287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.01,0.3812992095947266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,8,power_law_1.01,0.028275200724601747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.01,0.4367040157318115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,8,power_law_1.01,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.01,0.5856448173522949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,8,power_law_1.01,0.03266560137271881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.01,0.6801152229309082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,8,power_law_1.01,0.04069760143756866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.01,1.0012672424316407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,8,power_law_1.01,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.01,1.3055359840393066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,8,power_law_1.2,0.118995201587677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,8,power_law_1.01,0.05800319910049438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,8,power_law_1.01,1.6242559432983399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,8,power_law_1.2,0.1774656057357788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,8,power_law_1.01,0.06382079720497132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,8,power_law_1.01,2.2851583480834963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,8,power_law_1.2,0.09379199743270875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,8,power_law_1.01,0.07279360294342041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,8,power_law_1.2,0.12430720329284668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,8,power_law_1.01,0.07388799786567687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,8,power_law_1.01,4.472633743286133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,8,power_law_1.01,0.07936000227928161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,8,power_law_1.2,0.1486400008201599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,8,power_law_1.01,0.07895039916038513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,8,power_law_1.2,0.18506239652633666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,8,power_law_1.01,0.10655360221862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,8,power_law_1.2,0.2348992109298706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,8,power_law_1.01,0.1
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,8,power_law_1.2,0.24518399238586425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,8,power_law_1.01,0.09133440256118774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,8,power_law_1.2,0.2560447931289673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,8,power_law_1.01,0.14707839488983154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,8,power_law_1.2,0.2884927988052368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,8,power_law_1.01,0.1557760000228882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,8,power_law_1.2,0.27414400577545167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,8,power_law_1.01,0.15857919454574584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,8,power_law_1.2,0.27266559600830076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,8,power_law_1.01,0.15731199979782104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,8,power_law_1.2,0.2839423894882202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,8,power_law_1.2,0.28046720027923583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.01,0.15779199600219726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,8,power_law_1.2,0.3008512020111084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.01,0.15868159532546997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,8,power_law_1.2,0.3087615966796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.01,0.17297919988632202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,8,power_law_1.2,0.3131711959838867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.01,0.20232319831848145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,8,power_law_1.2,0.34817280769348147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.01,0.2301568031311035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,8,power_law_1.2,0.3521984100341797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.01,0.30944640636444093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,8,power_law_1.2,0.423635196685791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.01,0.374182391166687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,8,power_law_1.2,0.4755712032318115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.01,0.5288896083831787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,8,power_law_1.2,0.557919979095459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.01,0.6570496082305908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,8,power_law_1.2,0.6296127796173095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,8,power_law_1.01,0.8834367752075195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,8,power_law_1.2,0.8053695678710937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,8,power_law_1.01,1.382271957397461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,8,power_law_1.2,1.0865983963012695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,8,power_law_1.01,2.6458688735961915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,8,power_law_1.2,1.358835220336914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,8,power_law_1.2,1.8180416107177735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,8,power_law_1.2,2.065171241760254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,8,power_law_1.2,3.0800384521484374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,8,power_law_1.2,6.636243438720703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,8,power_law_1.2,0.09004160165786743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,8,power_law_1.2,0.12673920392990112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,8,power_law_1.2,0.07227519750595093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,8,power_law_1.2,0.09182720184326172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,8,power_law_1.2,0.107532799243927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,8,power_law_1.2,0.12244479656219483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,8,power_law_1.2,0.15444480180740355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,8,power_law_1.2,0.16261759996414185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,8,power_law_1.2,0.1647744059562683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,8,power_law_1.2,0.1705407977104187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,8,power_law_1.2,0.16836479902267457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,8,power_law_1.2,0.18401919603347777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,8,power_law_1.2,0.18428800106048585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,8,power_law_1.2,0.19063040018081664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,8,power_law_1.2,0.20130560398101807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,8,power_law_1.2,0.2022144079208374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,8,power_law_1.2,0.21107199192047119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,8,power_law_1.2,0.2265023946762085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,8,power_law_1.2,0.2522176027297974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,8,power_law_1.2,0.290828800201416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,8,power_law_1.2,0.34225919246673586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,8,power_law_1.2,0.442412805557251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,8,power_law_1.2,0.4988096237182617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,8,power_law_1.2,0.6749631881713867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,8,power_law_1.2,0.8273152351379395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,8,power_law_1.2,1.1261119842529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,8,power_law_1.2,0.10238080024719239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,8,power_law_1.2,1.5126655578613282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,8,power_law_1.2,0.1288831949234009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,8,power_law_1.2,2.033145523071289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,8,power_law_1.2,0.07623040080070495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,8,power_law_1.2,0.08956159949302674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,8,power_law_1.2,2.8641984939575194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,8,power_law_1.2,0.11198079586029053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,8,power_law_1.2,5.984998321533203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,8,power_law_1.2,0.13111679553985595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,8,power_law_1.2,0.16619520187377929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,8,power_law_1.2,0.17681280374526978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,8,power_law_1.2,0.17339520454406737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,8,power_law_1.2,0.18128639459609985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,8,power_law_1.2,0.1813055992126465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,8,power_law_1.2,0.18243839740753173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,8,power_law_1.2,0.1898751974105835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,8,power_law_1.2,0.18999680280685424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,8,power_law_1.2,0.1953727960586548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,8,power_law_1.2,0.2060159921646118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,8,power_law_1.2,0.19702399969100953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,8,power_law_1.2,0.23461759090423584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.2,0.25100159645080566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.2,0.27550079822540285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.2,0.29950718879699706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,8,power_law_1.2,0.028288000822067262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.2,0.39521279335021975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,8,power_law_1.2,0.048390400409698484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.2,0.45479679107666016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,8,power_law_1.2,0.030022400617599487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.2,0.618342399597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,8,power_law_1.2,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.2,0.7155712127685547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,8,power_law_1.2,0.043654400110244754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,8,power_law_1.2,0.053427201509475705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.2,1.0518400192260742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,4,balanced,0.06446933249632518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,4,balanced,0.07354666789372762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,8,power_law_1.2,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.2,1.3095680236816407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,4,balanced,0.09051733215649922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,4,balanced,0.12285332878430684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,8,power_law_1.2,0.06685439944267273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,8,power_law_1.2,1.6714176177978515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,4,balanced,0.18386665980021158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,8,power_law_1.2,0.07524480223655701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,4,balanced,0.3057120045026143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,8,power_law_1.2,2.6556800842285155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,4,balanced,0.30001066128412884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,4,balanced,0.3049333294232686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,8,power_law_1.2,4.967897415161133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,4,balanced,0.30268265803654987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,4,balanced,0.31121599674224854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,4,balanced,0.2972426613171895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,4,balanced,0.29315733909606934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,4,balanced,0.3076053261756897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,4,balanced,0.29733333985010785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,4,balanced,0.3073226610819499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,4,balanced,0.31305599212646484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,4,balanced,0.31361599763234455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,4,balanced,0.3386559883753459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,4,balanced,0.33472001552581787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,4,balanced,0.35761066277821857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,4,balanced,0.37436266740163165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,4,balanced,0.42166932423909503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,4,balanced,0.47089068094889325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,4,balanced,0.6454346577326456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,4,balanced,0.7272586822509766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,4,balanced,1.0263253053029378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,4,balanced,1.292741298675537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,4,balanced,0.056549335519472756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,4,balanced,0.0647573322057724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,4,balanced,0.06804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,4,balanced,0.09115733702977498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,4,balanced,0.12972799936930338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,4,balanced,0.20625599225362143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,4,balanced,0.19420800606409708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,4,balanced,0.19250667095184326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,4,balanced,0.19729600350062051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,4,balanced,0.19614932934443155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,4,balanced,1.556272029876709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,4,balanced,0.1978666583697001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,4,balanced,0.19620267550150552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,4,balanced,0.19926933447519937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,4,balanced,0.1988053321838379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,4,balanced,0.20497065782546997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,4,balanced,0.2111253341039022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,4,balanced,0.216213325659434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,4,balanced,0.2422773241996765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,4,balanced,0.2490773399670919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,4,balanced,0.2797546585400899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,4,balanced,0.30692267417907715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,4,balanced,0.3682560125986735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,4,balanced,2.408250649770101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,4,balanced,0.42956264813741046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,4,balanced,0.062447999914487205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,4,balanced,0.06933866441249847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,4,balanced,0.6130079825719198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,4,balanced,0.07980800171693166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,4,balanced,0.10662933190663655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,4,balanced,0.15660799543062845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,4,balanced,0.23366934061050415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,4,balanced,0.2254133423169454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,4,balanced,0.21977599461873373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,4,balanced,0.22030933698018393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,4,balanced,0.21819732586542764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,4,balanced,0.20630399386088052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,4,balanced,0.7286240259806315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,4,balanced,0.20705600579579672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,4,balanced,0.2120800018310547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,4,balanced,0.20761066675186157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,4,balanced,0.21625600258509317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,4,balanced,0.2271626591682434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,4,balanced,0.2216106653213501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,8,power_law_1.2,0.07403519749641418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,4,balanced,0.24048000574111938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,8,power_law_1.2,0.0786624014377594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,4,balanced,0.2495680054028829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,8,power_law_1.2,0.09779840111732482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,4,balanced,1.037338654200236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,8,power_law_1.2,0.09904000163078308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,4,balanced,0.2669066588083903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,8,power_law_1.2,0.09075199961662292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,8,power_law_1.2,0.1524608016014099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,4,balanced,0.28737600644429523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,8,power_law_1.2,0.15525120496749878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,8,power_law_1.2,0.16046080589294434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,8,power_law_1.2,0.15749119520187377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,4,balanced,0.335807998975118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.2,0.15348479747772217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.2,0.15955840349197387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.2,0.17539199590682983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.2,0.2124351978302002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,4,balanced,4.849018732706706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,4,balanced,0.38362665971120197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,4,balanced,1.3269866307576497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.2,0.2371583938598633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.2,0.3285248041152954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.2,0.4335936069488525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.2,0.5926271915435791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,4,balanced,0.5793439944585165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.2,0.7672895908355712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,8,power_law_1.2,0.9662528038024902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,8,power_law_1.2,1.4565631866455078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,8,power_law_1.2,2.945363235473633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,4,balanced,0.6658773422241211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,4,balanced,1.6185280481974285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,4,balanced,0.9333173433939616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,4,balanced,1.2154933611551921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,4,balanced,2.5126187006632485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,4,balanced,1.4852852821350098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,4,balanced,2.3060213724772134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,4,balanced,4.966927846272786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,4,balanced,0.023610666394233704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,4,balanced,0.02446399877468745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,4,balanced,0.028714666763941448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,4,balanced,0.039173332353432976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,4,balanced,0.05693333347638448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,4,balanced,0.08381866415341695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,4,balanced,0.08538132905960083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,4,balanced,0.08939733107884724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,4,balanced,0.09201066692670186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,4,balanced,0.0921013355255127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,4,power_law_1.01,0.09553279876708984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,4,balanced,0.09647466739018758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,4,balanced,0.10140267014503479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,4,power_law_1.01,0.12357759475708008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,4,balanced,0.10217066605885823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,4,power_law_1.01,0.09405440092086792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,4,balanced,0.10895466804504395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,4,power_law_1.01,0.1193343997001648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,4,balanced,0.1367093324661255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,4,power_law_1.01,0.16300159692764282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,4,balanced,0.13322133819262186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,4,power_law_1.01,0.1910591959953308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,4,balanced,0.14053866267204285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,4,power_law_1.01,0.24511361122131348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,4,balanced,0.20791999499003092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,4,power_law_1.01,0.2504832029342651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,4,balanced,4.521727879842122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,4,power_law_1.01,0.24814720153808595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,4,balanced,0.22393065690994263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,4,power_law_1.01,0.25824000835418703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,4,balanced,0.22895467281341553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,4,power_law_1.01,0.2618688106536865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,4,power_law_1.01,0.27331840991973877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,4,balanced,0.24569066365559897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,4,power_law_1.01,0.2657919883728027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,4,power_law_1.01,0.28499839305877683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,4,power_law_1.01,0.2976576089859009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,4,balanced,0.27668799956639606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,4,power_law_1.01,0.2900480031967163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,4,power_law_1.01,0.28672640323638915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,4,power_law_1.01,0.31880319118499756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,4,balanced,0.3059840003649394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,4,power_law_1.01,0.3382848024368286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,4,power_law_1.01,0.37598719596862795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,4,power_law_1.01,0.41877121925354005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,4,power_law_1.01,0.5165696144104004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,4,balanced,0.3718186616897583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,4,power_law_1.01,0.6095808029174805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,4,power_law_1.01,0.780460786819458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,4,power_law_1.01,0.9659520149230957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,4,power_law_1.01,1.3640512466430663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,4,balanced,0.43421868483225506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,4,power_law_1.01,1.6120384216308594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,4,power_law_1.01,2.0448320388793944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,4,power_law_1.01,3.0783615112304688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,4,power_law_1.01,6.223603057861328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,4,balanced,0.6871840159098307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,4,balanced,0.8262826601664225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,4,balanced,1.0811839898427327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,4,balanced,1.601685365041097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,4,power_law_1.01,0.07377279996871948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,4,power_law_1.01,0.09432960152626038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,4,power_law_1.01,0.07428479790687562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,4,power_law_1.01,0.08420479893684388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,4,power_law_1.01,0.11089919805526734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,4,power_law_1.01,0.12942719459533691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,4,power_law_1.01,0.1582335948944092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,4,power_law_1.01,0.15985920429229736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,4,power_law_1.01,0.1680448055267334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,4,power_law_1.01,0.1692863941192627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,4,power_law_1.01,0.17614079713821412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,4,power_law_1.01,0.1788800001144409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,4,power_law_1.01,0.1827712059020996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,4,power_law_1.01,0.18481279611587526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,4,power_law_1.01,0.19595520496368407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,4,power_law_1.01,0.20003199577331543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,4,power_law_1.01,0.20428800582885742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,4,power_law_1.01,0.23568639755249024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,4,power_law_1.01,0.25197439193725585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,4,power_law_1.01,0.08343679904937744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,4,power_law_1.01,0.3004096031188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,4,power_law_1.01,0.1047808051109314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,4,power_law_1.01,0.34126079082489014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,4,power_law_1.01,0.08289920091629029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,4,power_law_1.01,0.42920961380004885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,4,power_law_1.01,0.09879680275917054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,4,power_law_1.01,0.4947968006134033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,4,power_law_1.01,0.12319999933242798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,4,power_law_1.01,0.6699071884155273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,4,power_law_1.01,0.15372159481048583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,4,power_law_1.01,0.8522175788879395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,4,power_law_1.01,0.17523200511932374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,4,power_law_1.01,1.2575872421264649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,4,power_law_1.01,0.17772159576416016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,4,power_law_1.01,1.613702392578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,4,power_law_1.01,0.18431999683380126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,4,balanced,3.1436214447021484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,4,power_law_1.01,1.8314943313598633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,4,power_law_1.01,0.19475200176239013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,4,power_law_1.01,0.19377919435501098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,4,power_law_1.01,2.842905616760254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,4,power_law_1.01,0.19605120420455932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,4,power_law_1.01,0.022892799973487855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,4,power_law_1.01,5.982828903198242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,4,power_law_1.01,0.2029184103012085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,4,power_law_1.01,0.02917119860649109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,4,power_law_1.01,0.2021183967590332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,4,power_law_1.01,0.029344001412391664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,4,power_law_1.01,0.20610558986663818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,4,power_law_1.01,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,4,power_law_1.01,0.21754879951477052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,4,power_law_1.01,0.04679040014743805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,4,power_law_1.01,0.21719040870666503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,4,power_law_1.01,0.05299839973449707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,4,power_law_1.01,0.24039039611816407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,4,power_law_1.01,0.06792960166931153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.01,0.2536895990371704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,4,power_law_1.01,0.06915199756622314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.01,0.29130239486694337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,4,power_law_1.01,0.06970880031585694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.01,0.3408447980880737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,4,power_law_1.01,0.07370240092277527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.01,0.4157951831817627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,4,power_law_1.01,0.07796480059623719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.01,0.5072447776794433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,4,power_law_1.01,0.08072959780693054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.01,0.6704063892364502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,4,power_law_1.01,0.07972480058670044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.01,0.8400064468383789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,4,power_law_1.01,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.01,1.1785344123840331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,4,power_law_1.01,0.09434880018234253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,4,power_law_1.01,0.09103999733924865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.01,1.4089471817016601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,4,power_law_1.01,0.09297919869422913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,4,power_law_1.01,1.7528127670288085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,4,power_law_1.01,0.1700543999671936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,4,power_law_1.01,2.690719985961914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.01,0.1880959987640381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.01,0.17072000503540039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,4,power_law_1.01,5.861721420288086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.01,0.1829632043838501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.01,0.20728960037231445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.01,0.24252800941467284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.01,0.31240320205688477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.01,0.36932480335235596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.01,0.5109951972961426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.01,0.6318079948425293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,4,power_law_1.01,0.7693312168121338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,4,power_law_1.01,1.2221055984497071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,4,power_law_1.01,2.3269439697265626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,4,power_law_1.2,0.09283199906349182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,4,power_law_1.2,0.11610239744186401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,4,power_law_1.2,0.08815360069274902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,4,power_law_1.2,0.0753920018672943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,4,power_law_1.2,0.10728319883346557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,4,power_law_1.2,0.09311360120773315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,4,power_law_1.2,0.1548095941543579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,4,power_law_1.2,0.06767359972000123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,4,power_law_1.2,0.08680319786071777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,4,power_law_1.2,0.17381759881973266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,4,power_law_1.2,0.11165440082550049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,4,power_law_1.2,0.23155839443206788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,4,power_law_1.2,0.12343039512634277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,4,power_law_1.2,0.24180479049682618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,4,power_law_1.2,0.16631679534912108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,4,power_law_1.2,0.24718079566955567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,4,power_law_1.2,0.1637120008468628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,4,power_law_1.2,0.2587968111038208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,4,power_law_1.2,0.16825599670410157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,4,power_law_1.2,0.26903040409088136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,4,power_law_1.2,0.174726402759552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,4,power_law_1.2,0.2686847925186157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,4,power_law_1.2,0.17274880409240723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,4,power_law_1.2,0.2621567964553833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,4,power_law_1.2,0.17724159955978394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,4,power_law_1.2,0.28218879699707033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,4,power_law_1.2,0.18553600311279297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,4,power_law_1.2,0.29533441066741944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,4,power_law_1.2,0.192249596118927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,4,power_law_1.2,0.30655999183654786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,4,power_law_1.2,0.20307838916778564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,4,power_law_1.2,0.3168384075164795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,4,power_law_1.2,0.2016767978668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,4,power_law_1.2,0.3328448057174683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,4,power_law_1.2,0.22391040325164796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,4,power_law_1.2,0.3559743881225586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,4,power_law_1.2,0.23558399677276612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,4,power_law_1.2,0.40894079208374023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,4,power_law_1.2,0.26513919830322263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,4,power_law_1.2,0.4513088226318359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,4,power_law_1.2,0.3137727975845337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,4,power_law_1.2,0.5303296089172364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,4,power_law_1.2,0.6139135837554932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,4,power_law_1.2,0.3533375978469849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,4,power_law_1.2,0.4497471809387207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,4,power_law_1.2,0.8161855697631836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,4,power_law_1.2,0.5337344169616699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,4,power_law_1.2,1.0413375854492188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,4,power_law_1.2,0.08163840174674988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,4,power_law_1.2,0.7341055870056152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,4,power_law_1.2,0.10859520435333252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,4,power_law_1.2,1.3205632209777831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,4,power_law_1.2,0.9590720176696778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,4,power_law_1.2,0.07608320116996765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,4,power_law_1.2,1.6944639205932617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,4,power_law_1.2,0.09094399809837342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,4,power_law_1.2,1.329030418395996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,4,power_law_1.2,2.0812416076660156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,4,power_law_1.2,0.11835520267486573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,4,power_law_1.2,1.7496000289916993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,4,power_law_1.2,3.896051025390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,4,power_law_1.2,0.1372928023338318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,4,power_law_1.2,2.03121280670166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,4,power_law_1.2,0.1708799958229065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,4,power_law_1.2,8.14824981689453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,4,power_law_1.2,3.294803237915039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,4,power_law_1.2,0.1938431978225708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,4,power_law_1.2,0.19173120260238646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,4,power_law_1.2,5.986003112792969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,4,power_law_1.2,0.186188805103302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,4,power_law_1.2,0.200927996635437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,4,power_law_1.2,0.2066431999206543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,4,power_law_1.2,0.022431999444961548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,4,power_law_1.2,0.20551679134368897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,4,power_law_1.2,0.027609598636627198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,4,power_law_1.2,0.20691840648651122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,4,power_law_1.2,0.02794879972934723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,4,power_law_1.2,0.21328001022338866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,4,power_law_1.2,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,4,power_law_1.2,0.22402560710906982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,4,power_law_1.2,0.046188798546791074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,4,power_law_1.2,0.2241152048110962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,4,power_law_1.2,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,4,power_law_1.2,0.24447360038757324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,4,power_law_1.2,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.2,0.2769599914550781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,4,power_law_1.2,0.0705407977104187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.2,0.3041919946670532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,4,power_law_1.2,0.07278720140457154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.2,0.36369919776916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,4,power_law_1.2,0.07268480062484742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.2,0.45102720260620116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,4,power_law_1.2,0.0781823992729187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.2,0.5151040077209472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,4,power_law_1.2,0.07771520018577575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.2,0.6943871974945068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,4,power_law_1.2,0.08309760093688964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.2,0.8658623695373535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,4,power_law_1.2,0.08599039912223816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.2,1.1949119567871094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,4,power_law_1.2,0.09809280037879944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.2,1.5399231910705566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,4,power_law_1.2,0.09641600251197815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,4,power_law_1.2,1.8621183395385743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,4,power_law_1.2,0.09854080080986023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,4,power_law_1.2,2.7330816268920897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,4,power_law_1.2,0.17454080581665038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.2,0.19759999513626098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,4,power_law_1.2,5.986393737792969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.2,0.17685120105743407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.2,0.19218560457229614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.2,0.21791360378265381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.2,0.2568703889846802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.2,0.3218111991882324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.2,0.4044544219970703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.2,0.533516788482666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.2,0.7114496231079102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,4,power_law_1.2,0.7772928237915039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,4,power_law_1.2,1.2394304275512695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,4,power_law_1.2,2.520908737182617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,2,balanced,0.06881600121657054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,2,balanced,0.06187200049559275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,2,balanced,0.0842026670773824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,2,balanced,0.06592000027497609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,2,balanced,0.10036266843477885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,2,balanced,0.07828799883524577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,2,balanced,0.13109333316485086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,2,balanced,0.10346133510271709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,2,balanced,0.1992266575495402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,2,balanced,0.15235199530919394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,2,balanced,0.3322346607844035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,2,balanced,0.23549334208170572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,2,balanced,0.33194132645924884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,2,balanced,0.2283946673075358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,2,balanced,0.320848007996877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,2,balanced,0.22953067223230997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,2,balanced,0.32286399602890015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,2,balanced,0.2295359969139099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,2,balanced,0.3208426634470622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,2,balanced,0.22235733270645142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,2,balanced,0.3201013406117757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,2,balanced,0.22612265745798746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,2,balanced,0.3163999915122986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,2,balanced,0.22873600323994955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,2,balanced,0.3232480088869731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,2,balanced,0.23040533065795898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,2,balanced,0.3248586654663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,2,balanced,0.23519466320673624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,2,balanced,0.3234826723734538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,2,balanced,0.24414400259653726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,2,balanced,0.33947734038035077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,2,balanced,0.24823999404907227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,2,balanced,0.34041066964467365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,2,balanced,0.2579999963442485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,2,balanced,0.3614613215128581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,2,balanced,0.28693334261576336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,2,balanced,0.3829066753387451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,2,balanced,0.30292266607284546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,2,balanced,0.41602134704589844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,2,balanced,0.3505653142929077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,2,balanced,0.45134933789571124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,2,balanced,0.3918986717859904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,2,balanced,0.5311520099639893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,2,balanced,0.48608001073201496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,2,balanced,0.07296533385912578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,2,balanced,0.607040007909139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,2,balanced,0.07982400059700012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,2,balanced,0.09237333138783772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,2,balanced,0.12828800082206726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,2,balanced,0.5759093364079794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,2,balanced,0.18665599822998047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,2,balanced,0.2826720078786214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,2,balanced,0.8643413384755453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,2,balanced,0.2774453361829122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,2,balanced,0.2653013269106547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,2,balanced,0.2712373336156209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,2,balanced,0.8293759822845459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,2,balanced,0.2664533257484436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,2,balanced,0.2677119970321655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,2,balanced,0.2687679926554362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,2,balanced,0.023605334262053173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,2,balanced,0.27111999193827313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,2,balanced,0.02553066611289978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,2,balanced,1.0119946797688801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,2,balanced,0.028698667883872986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,2,balanced,0.27662932872772217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,2,balanced,0.036576000352700554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,2,balanced,1.0152746836344402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,2,balanced,0.2847306728363037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,2,balanced,0.0611413319905599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,2,balanced,0.2916319966316223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,2,balanced,0.0839573343594869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,2,balanced,0.08573333422342937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,2,balanced,0.2983679970105489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,2,balanced,0.08988266189893086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,2,balanced,0.0934986670811971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,2,balanced,0.3214293320973714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,2,balanced,0.09707199533780415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,2,balanced,1.4235520362854004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,2,balanced,0.10314133763313293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,2,balanced,0.33393601576487225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,2,balanced,0.10099732875823975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,2,balanced,0.10044266780217488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,2,balanced,1.4574400583902996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,2,balanced,0.3701813220977783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,2,balanced,0.11000532905260722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,2,balanced,0.11179733276367188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,2,balanced,0.40004265308380127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,2,balanced,0.11408533652623494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,2,balanced,0.12174933155377705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,2,balanced,0.14619732896486917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,2,balanced,0.4853173494338989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,2,balanced,0.1702666680018107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,2,balanced,1.8056693077087402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,2,balanced,0.24736533562342325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,2,balanced,1.8698347409566243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,2,balanced,0.5622986555099487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,2,balanced,0.265882670879364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,2,balanced,0.3022666573524475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,2,balanced,0.842357317606608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,2,balanced,0.3273013234138489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,2,balanced,2.1801973978678384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,2,balanced,0.3762933413187663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,2,balanced,2.290250619252523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,2,balanced,0.9949759642283121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,2,balanced,0.41650664806365967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,2,balanced,1.423749287923177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,2,balanced,0.682474692662557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,2,balanced,3.384079933166504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,2,balanced,3.5572532018025718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,2,balanced,0.7898773352305094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,2,balanced,1.8460480372111003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,2,balanced,1.0548266569773357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,2,balanced,2.275338649749756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,2,balanced,1.5432693163553874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,2,balanced,6.590378443400065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,2,balanced,7.011578877766927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,2,balanced,3.5525118509928384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,2,balanced,3.032400131225586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,2,balanced,6.986421585083008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,2,power_law_1.01,0.0828607976436615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,2,power_law_1.01,0.09986559748649597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,2,power_law_1.01,0.09228159785270691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,2,power_law_1.01,0.11080319881439209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,2,power_law_1.01,0.0670144021511078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,2,power_law_1.01,0.1674623966217041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,2,power_law_1.01,0.08095359802246094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,2,power_law_1.01,0.1853503942489624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,2,power_law_1.01,0.07267199754714966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,2,power_law_1.01,0.24569599628448485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,2,power_law_1.01,0.10156160593032837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,2,power_law_1.01,0.2475584030151367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,2,power_law_1.01,0.12878079414367677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,2,power_law_1.01,0.2534143924713135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,2,power_law_1.01,0.14573440551757813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,2,power_law_1.01,0.2587584018707275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,2,power_law_1.01,0.18159359693527222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,2,power_law_1.01,0.2675136089324951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,2,power_law_1.01,0.19118080139160157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,2,power_law_1.01,0.18888319730758668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,2,power_law_1.01,0.2734272003173828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,2,power_law_1.01,0.19534080028533934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,2,power_law_1.01,0.27838079929351806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,2,power_law_1.01,0.20398080348968506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,2,power_law_1.01,0.28650240898132323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,2,power_law_1.01,0.20373759269714356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,2,power_law_1.01,0.2993727922439575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,2,power_law_1.01,0.210265588760376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,2,power_law_1.01,0.3033663988113403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,2,power_law_1.01,0.2163327932357788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,2,power_law_1.01,0.3068543910980225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,2,power_law_1.01,0.22656641006469727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,2,power_law_1.01,0.34535679817199705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,2,power_law_1.01,0.3754879951477051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,2,power_law_1.01,0.4186560153961182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,2,power_law_1.01,0.4725696086883545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,2,power_law_1.01,0.6099391937255859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,2,power_law_1.01,0.6938047885894776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,2,power_law_1.01,0.897760009765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,2,power_law_1.01,1.1102911949157714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,2,power_law_1.01,1.5937600135803223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,2,power_law_1.01,1.985856056213379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,2,power_law_1.01,2.4464895248413088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,2,power_law_1.01,3.69128303527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,2,power_law_1.01,7.167603302001953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,2,power_law_1.01,0.07834240198135375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,2,power_law_1.01,0.09418879747390747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,2,power_law_1.01,0.09236479997634887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,2,power_law_1.01,0.11465599536895751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,2,power_law_1.01,0.15375360250473022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,2,power_law_1.01,0.16475520133972169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,2,power_law_1.01,0.2174976110458374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,2,power_law_1.01,0.22600319385528564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,2,power_law_1.01,0.22697598934173585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,2,power_law_1.01,0.2283519983291626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,2,power_law_1.01,0.24330239295959472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,2,power_law_1.01,0.2476736068725586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,2,power_law_1.01,0.25189759731292727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,2,power_law_1.01,0.02133760005235672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,2,power_law_1.01,0.2563391923904419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,2,power_law_1.01,0.025798401236534117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,2,power_law_1.01,0.2661504030227661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,2,power_law_1.01,0.026009601354599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,2,power_law_1.01,0.26956160068511964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,2,power_law_1.01,0.03307519853115082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,2,power_law_1.01,0.27781760692596436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,2,power_law_1.01,0.05103359818458557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,2,power_law_1.01,0.3099967956542969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,2,power_law_1.01,0.05751680135726929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.01,0.3368767976760864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,2,power_law_1.01,0.06776319742202759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.01,0.3848639965057373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,2,power_law_1.01,0.07265920042991639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.01,0.4456064224243164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,2,power_law_1.01,0.07744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.01,0.5540544033050537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,2,power_law_1.01,0.08150399923324585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.01,0.6990592002868652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,2,power_law_1.01,0.08604800105094909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.01,0.9093888282775879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,2,power_law_1.01,0.08362240195274354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.01,1.1165056228637695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,2,power_law_1.01,0.08598399758338929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.01,1.5572928428649901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,2,power_law_1.01,0.09455360174179077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.01,2.0422527313232424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,2,power_law_1.01,0.0907904028892517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,2,power_law_1.01,2.533919906616211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,2,power_law_1.01,0.09443839788436889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,2,power_law_1.01,0.10364799499511719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,2,power_law_1.01,3.747148895263672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,2,power_law_1.01,0.11886080503463745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,2,power_law_1.01,7.274905395507813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.01,0.13110400438308717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.01,0.22025599479675292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.01,0.2445120096206665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.01,0.24493439197540284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.01,0.2752000093460083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.01,0.3369920015335083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.01,0.40500478744506835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.01,0.5478079795837403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.01,0.687391996383667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,2,power_law_1.01,0.8278079986572265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,2,power_law_1.01,1.2761792182922362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,2,power_law_1.01,2.474188804626465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,2,power_law_1.2,0.08291839957237243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,2,power_law_1.2,0.09916800260543823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,2,power_law_1.2,0.0882752001285553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,2,power_law_1.01,0.230131196975708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,2,power_law_1.2,0.11593600511550903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,2,power_law_1.01,0.24349439144134521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,2,power_law_1.2,0.1477120041847229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,2,power_law_1.01,0.2805504083633423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,2,power_law_1.01,0.30396161079406736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,2,power_law_1.2,0.17620480060577393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,2,power_law_1.01,0.35891199111938477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,2,power_law_1.2,0.23958399295806884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,2,power_law_1.2,0.24833920001983642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,2,power_law_1.01,0.4146431922912598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,2,power_law_1.01,0.5404928207397461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,2,power_law_1.2,0.2559295892715454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,2,power_law_1.2,0.2660288095474243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,2,power_law_1.01,0.6463871955871582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,2,power_law_1.2,0.2631103992462158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,2,power_law_1.01,0.8904383659362793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,2,power_law_1.2,0.2747136116027832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,2,power_law_1.01,1.1125056266784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,2,power_law_1.2,0.28066558837890626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,2,power_law_1.01,1.4950207710266112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,2,power_law_1.2,0.29377920627593995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,2,power_law_1.01,1.990604782104492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,2,power_law_1.2,0.30916481018066405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,2,power_law_1.01,2.50512638092041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,2,power_law_1.2,0.3073728084564209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,2,power_law_1.2,0.3372351884841919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,2,power_law_1.01,3.817241668701172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,2,power_law_1.2,0.3664191961288452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,2,power_law_1.2,0.3845632076263428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,2,power_law_1.01,7.6424705505371096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,2,power_law_1.2,0.4496767997741699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,2,power_law_1.2,0.5004032135009766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,2,power_law_1.2,0.6211904048919678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,2,power_law_1.2,0.7133056163787842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,2,power_law_1.2,0.9502911567687988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,2,power_law_1.2,1.124019241333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,2,power_law_1.2,1.581606388092041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,2,power_law_1.2,1.9930047988891602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,2,power_law_1.2,2.4908031463623046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,2,power_law_1.2,3.7559871673583984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,2,power_law_1.2,7.977945709228516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,2,power_law_1.2,0.06881279945373535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,2,power_law_1.2,0.08093439936637878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,2,power_law_1.2,0.07408639788627625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,2,power_law_1.2,0.09807999730110169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,2,power_law_1.2,0.11210240125656128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,2,power_law_1.2,0.14610559940338136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,2,power_law_1.2,0.1738495945930481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,2,power_law_1.2,0.1881343960762024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,2,power_law_1.2,0.19389439821243287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,2,power_law_1.2,0.1965824007987976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,2,power_law_1.2,0.20051200389862062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,2,power_law_1.2,0.21553919315338135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,2,power_law_1.2,0.21198720932006837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,2,power_law_1.2,0.21750400066375733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,2,power_law_1.2,0.22858240604400634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,2,power_law_1.2,0.24817919731140137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,2,power_law_1.2,0.24569599628448485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,2,power_law_1.2,0.28716800212860105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,2,power_law_1.2,0.31948161125183105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,2,power_law_1.2,0.07761279940605163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,2,power_law_1.2,0.3717888116836548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,2,power_law_1.2,0.09470080137252808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,2,power_law_1.2,0.42485761642456055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,2,power_law_1.2,0.08762879967689514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,2,power_law_1.2,0.5643072128295898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,2,power_law_1.2,0.1109887957572937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,2,power_law_1.2,0.6613952159881592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,2,power_law_1.2,0.13664000034332274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,2,power_law_1.2,0.8651455879211426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,2,power_law_1.2,0.16022399663925171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,2,power_law_1.2,1.130732822418213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,2,power_law_1.2,0.2091775894165039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,2,power_law_1.2,0.22288000583648682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,2,power_law_1.2,1.6599296569824218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,2,power_law_1.2,0.2259648084640503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,2,power_law_1.2,2.1730815887451174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,2,power_law_1.2,0.2264256000518799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,2,power_law_1.2,2.5776960372924806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,2,power_law_1.2,0.2363840103149414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,2,power_law_1.2,4.004940795898437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,2,power_law_1.2,0.24902400970458985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,2,power_law_1.2,0.24832639694213868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,2,power_law_1.2,7.873683166503906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,2,power_law_1.2,0.26026880741119385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,2,power_law_1.2,0.2751744031906128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,2,power_law_1.2,0.2825279951095581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,2,power_law_1.2,0.28886399269104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,2,power_law_1.2,0.3236095905303955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.2,0.3530751943588257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.2,0.40303359031677244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.2,0.46576638221740724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.2,0.622105598449707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.2,0.7212672233581543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.2,0.9439616203308105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.2,1.1665663719177246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.2,1.6782655715942383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.2,2.00775032043457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,2,power_law_1.2,2.5598400115966795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,2,power_law_1.2,3.9841983795166014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,2,power_law_1.2,8.192473602294921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,2,power_law_1.2,0.021350400149822236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,2,power_law_1.2,0.024736000597476958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,2,power_law_1.2,0.026572799682617186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,2,power_law_1.2,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,2,power_law_1.2,0.04792959988117218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,2,power_law_1.2,0.05482879877090454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,2,power_law_1.2,0.0680895984172821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,2,power_law_1.2,0.07191039919853211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,2,power_law_1.2,0.07615360021591186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,1,balanced,0.07935466865698497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,1,balanced,0.09412800272305806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,2,power_law_1.2,0.07946879863739013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,1,balanced,0.11642666657765706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,2,power_law_1.2,0.08568959832191467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,1,balanced,0.15680533647537231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,2,power_law_1.2,0.0830847978591919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,1,balanced,0.2339466611544291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,1,balanced,0.4048853317896525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,2,power_law_1.2,0.08728320002555848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,1,balanced,0.3980853160222371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,2,power_law_1.2,0.0945792019367218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,1,balanced,0.3877439896265666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,1,balanced,0.3882720073064168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,2,power_law_1.2,0.0927616000175476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,1,balanced,0.38754133383433026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,2,power_law_1.2,0.09959040284156799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,1,balanced,0.3880266745885213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,2,power_law_1.2,0.1055232048034668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,1,balanced,0.39452266693115234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,1,balanced,0.39937599500020343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,2,power_law_1.2,0.12382080554962158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,1,balanced,0.40493865807851154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.2,0.1365056037902832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,1,balanced,0.4103093147277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.2,0.2340480089187622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,1,balanced,0.4174400170644124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.2,0.25182719230651857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,1,balanced,0.43011200428009033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.2,0.24829440116882323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.2,0.2778239965438843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,1,balanced,0.4700106779734294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.2,0.35786240100860595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,1,balanced,0.4971359968185425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.2,0.42671999931335447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,1,balanced,0.5684373378753662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.2,0.5619711875915527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.2,0.7450496196746826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,1,balanced,0.625327984491984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,2,power_law_1.2,0.8552895545959472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,2,power_law_1.2,1.3071807861328124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,1,balanced,0.7578612963358561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,2,power_law_1.2,2.4829952239990236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,1,balanced,0.8915733496348063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,1,balanced,1.2964373429616292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,1,balanced,1.576485315958659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,1,balanced,0.07292266686757405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,1,balanced,0.07925333579381307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,1,balanced,0.09654933214187622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,1,balanced,0.1320746640364329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,1,balanced,0.18405866622924805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,1,balanced,0.2996266682942708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,1,balanced,2.297429402669271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,1,balanced,0.2950986623764038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,1,balanced,0.2986133297284444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,1,balanced,0.2999359965324402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,1,balanced,0.30161599318186444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,1,balanced,0.3039039969444275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,1,balanced,0.3118666609128316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,1,balanced,0.31434667110443115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,1,balanced,0.31942933797836304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,1,balanced,0.33187733093897503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,1,balanced,0.33609600861867267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,1,balanced,2.9131946563720703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,1,balanced,0.3504319985707601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,1,balanced,0.3945866823196411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,1,balanced,0.42475199699401855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,1,balanced,0.4973440170288086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,1,balanced,0.5600159962972006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,1,balanced,0.7066773573557535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,1,balanced,3.555002530415853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,1,balanced,0.10102933645248413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,1,balanced,0.11430933078130086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,1,balanced,0.1388266682624817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,1,balanced,0.8483520348866781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,1,balanced,0.18388267358144125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,1,balanced,0.27266667286554974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,1,balanced,0.455077330271403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,1,balanced,0.4479893445968628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,1,balanced,0.4540640115737915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,1,balanced,0.44494398434956867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,1,balanced,1.2478400071461995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,1,balanced,0.4302186568578084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,1,balanced,0.4315306742986043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,1,balanced,0.43833065032958984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,1,balanced,0.4352480173110962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,1,balanced,0.44285865624745685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,1,balanced,1.5265013376871746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,1,balanced,0.4489440123240153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,1,balanced,0.45501867930094403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,1,balanced,0.46773334344228107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,1,balanced,0.5113546848297119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,1,balanced,5.448042551676433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,1,balanced,0.5415893395741781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,1,balanced,2.2275999387105307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,1,balanced,0.6175200144449869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,1,balanced,0.6778026421864828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,1,balanced,0.8353386720021566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,1,balanced,2.901989301045736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,1,balanced,0.9814026355743408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,1,balanced,1.4863413174947102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,1,balanced,3.569178581237793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,1,balanced,1.7962239583333333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,1,balanced,0.023445333043734234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,1,balanced,0.024586667617162068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,1,balanced,0.026757332185904186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,1,balanced,0.03250666707754135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,1,balanced,0.046485334634780884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,1,balanced,0.07755200068155925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,1,balanced,0.07949866851170857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,1,balanced,0.0841919978459676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,1,balanced,0.08591467142105103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,1,balanced,10.584911982218424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,1,balanced,0.08742400010426839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,1,balanced,2.5734880765279136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,1,balanced,0.0916319986184438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,1,balanced,0.09794132908185323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,1,balanced,0.10294933120409648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,1,balanced,0.11428266763687134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,1,balanced,0.10601600011189778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,1,balanced,0.11032000184059143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,1,balanced,5.5859629313151045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,1,balanced,0.12455999851226807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,1,balanced,0.1249066690603892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,1,balanced,0.14053866267204285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,1,balanced,0.16710400581359863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,1,balanced,3.3740212122599282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,1,balanced,0.186463991800944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,1,balanced,0.24194665749867758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,1,balanced,0.29812800884246826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,1,balanced,0.33992000420888263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,1,balanced,4.169274648030599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,1,balanced,0.3864106734593709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,1,balanced,0.6085226535797119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,1,balanced,10.989023844401041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,1,balanced,0.707530657450358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,1,balanced,6.520325342814128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,1,balanced,0.9291466871897379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,1,balanced,1.3912960688273113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,1,power_law_1.01,0.07331200242042542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,1,power_law_1.01,0.08569599986076355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,1,power_law_1.01,0.09710080027580262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,1,power_law_1.01,0.13494399785995484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,1,power_law_1.01,0.1909119963645935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,1,power_law_1.01,0.21332480907440185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,1,power_law_1.01,0.2879487991333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,1,power_law_1.01,0.2986432075500488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,1,balanced,12.796122233072916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,1,power_law_1.01,0.30697600841522216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,1,power_law_1.01,0.30478079319000245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,1,power_law_1.01,0.32048640251159666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,1,power_law_1.01,0.3319231986999512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,1,power_law_1.01,0.3387775897979736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,1,power_law_1.01,0.34832000732421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,1,power_law_1.01,0.3571903944015503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,1,power_law_1.01,0.3701375961303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,1,power_law_1.01,0.38789119720458987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,1,power_law_1.01,0.44544639587402346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,1,power_law_1.01,0.4808512210845947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,1,power_law_1.01,0.5666431903839111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,1,power_law_1.01,0.6282495975494384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,1,power_law_1.01,0.8023360252380372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,1,power_law_1.01,0.9544063568115234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,1,power_law_1.01,1.2714816093444825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,1,power_law_1.01,1.5894399642944337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,1,power_law_1.01,2.2147903442382812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,1,balanced,2.7241973876953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,1,power_law_1.01,2.8418367385864256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,1,power_law_1.01,3.4921470642089845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,1,power_law_1.01,0.06675840020179749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,1,power_law_1.01,5.403430557250976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,1,power_law_1.01,0.07432320117950439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,1,power_law_1.01,0.08455680012702942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,1,power_law_1.01,10.557158660888671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,1,power_law_1.01,0.11499520540237426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,1,power_law_1.01,0.15427839756011963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,1,power_law_1.01,0.1723199963569641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,1,power_law_1.01,0.22609279155731202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,1,power_law_1.01,0.23956480026245117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,1,power_law_1.01,0.24131839275360106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,1,power_law_1.01,0.24798719882965087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,1,power_law_1.01,0.2591104030609131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,1,power_law_1.01,0.2682624101638794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,1,power_law_1.01,0.27317759990692136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,1,power_law_1.01,0.2837183952331543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,1,power_law_1.01,0.29993600845336915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,1,power_law_1.01,0.30794880390167234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,1,power_law_1.01,0.32198400497436525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,1,power_law_1.01,0.3732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,1,power_law_1.01,0.41408638954162597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,1,power_law_1.01,0.5046207904815674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,1,power_law_1.01,0.5837823867797851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,1,power_law_1.01,0.7514048099517823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,1,power_law_1.01,0.9172351837158204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,1,power_law_1.01,1.2516160011291504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,1,power_law_1.01,1.5821760177612305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,1,power_law_1.01,2.248396873474121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,1,power_law_1.01,2.9082496643066404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,1,power_law_1.01,3.5975616455078123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,1,power_law_1.01,5.602694320678711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,1,power_law_1.01,11.004096221923827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,1,power_law_1.01,0.0956928014755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,1,power_law_1.01,0.10826239585876465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,1,power_law_1.01,0.11660159826278686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,1,power_law_1.01,0.16325759887695312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,1,power_law_1.01,0.20864639282226563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,1,power_law_1.01,0.25144319534301757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,1,power_law_1.01,0.33087360858917236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,1,power_law_1.01,0.3369920015335083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,1,power_law_1.01,0.35228800773620605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,1,power_law_1.01,0.3535680055618286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,1,power_law_1.01,0.362009596824646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,1,power_law_1.01,0.3735296010971069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,1,power_law_1.01,0.3781440019607544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,1,power_law_1.01,0.3860608100891113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,1,power_law_1.01,0.40233597755432127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,1,power_law_1.01,0.41336960792541505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,1,power_law_1.01,0.4290175914764404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,1,power_law_1.01,0.493120002746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.01,0.5275519847869873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.01,0.6294464111328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,1,power_law_1.01,0.02139520049095154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.01,0.7248064041137695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,1,power_law_1.01,0.02285439968109131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.01,0.9227264404296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,1,power_law_1.01,0.02545279860496521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.01,1.1140992164611816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,1,power_law_1.01,0.031251201033592226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.01,1.507737636566162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,1,power_law_1.2,0.07284479737281799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,1,power_law_1.01,0.04613119959831238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,1,power_law_1.2,0.08443520069122315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.01,1.8976448059082032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,1,power_law_1.01,0.0686784029006958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,1,power_law_1.2,0.09715840220451355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.01,2.6710975646972654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,1,power_law_1.01,0.08046720027923585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,1,power_law_1.2,0.12691839933395385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.01,3.441209411621094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,1,power_law_1.01,0.08627840280532836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,1,power_law_1.2,0.1627776026725769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,1,power_law_1.01,0.08710399866104127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.01,4.2207489013671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,1,power_law_1.2,0.20563199520111083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,1,power_law_1.01,0.09030399918556213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,1,power_law_1.2,0.27626240253448486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,1,power_law_1.01,6.532288360595703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,1,power_law_1.01,0.10110080242156982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,1,power_law_1.2,0.2965759992599487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,1,power_law_1.01,0.10565760135650634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,1,power_law_1.01,12.722828674316407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,1,power_law_1.2,0.2986432075500488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,1,power_law_1.01,0.11070079803466797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,1,power_law_1.01,0.12273279428482056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,1,power_law_1.2,0.30368640422821047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,1,power_law_1.01,0.11521919965744018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,1,power_law_1.2,0.31841919422149656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,1,power_law_1.01,0.12435840368270874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,1,power_law_1.2,0.3294464111328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,1,power_law_1.01,0.13975679874420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,1,power_law_1.2,0.33583359718322753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,1,power_law_1.01,0.14113919734954833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,1,power_law_1.2,0.35618560314178466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.01,0.1583359956741333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,1,power_law_1.2,0.37450881004333497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.01,0.2003391981124878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,1,power_law_1.2,0.3884927988052368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.01,0.21695361137390137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,1,power_law_1.2,0.4023935794830322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.01,0.3144128084182739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,1,power_law_1.2,0.4636415958404541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.01,0.37115519046783446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,1,power_law_1.2,0.5008063793182373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.01,0.4131584167480469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,1,power_law_1.2,0.5934336185455322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,1,power_law_1.2,0.6658751964569092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.01,0.49027199745178224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,1,power_law_1.2,0.8206015586853027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.01,0.6490367889404297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,1,power_law_1.2,0.9695615768432617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.01,0.804748821258545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,1,power_law_1.2,1.2753087997436523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.01,0.9712960243225097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,1,power_law_1.2,1.5950847625732423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,1,power_law_1.01,1.4715264320373536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,1,power_law_1.2,2.22225284576416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,1,power_law_1.01,2.798201560974121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,1,power_law_1.2,2.8626495361328126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,1,power_law_1.2,3.4921409606933596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,1,power_law_1.2,0.06885120272636414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,1,power_law_1.2,0.07544320225715637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,1,power_law_1.2,5.41580810546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,1,power_law_1.2,0.08369920253753663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,1,power_law_1.2,0.11288319826126099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,1,power_law_1.2,10.540729522705078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,1,power_law_1.2,0.13189760446548462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,1,power_law_1.2,0.1665984034538269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,1,power_law_1.2,0.2245311975479126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,1,power_law_1.2,0.24028799533843995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,1,power_law_1.2,0.2445120096206665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,1,power_law_1.2,0.2483135938644409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,1,power_law_1.2,0.259168004989624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,1,power_law_1.2,0.2670655965805054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,1,power_law_1.2,0.2785856008529663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,1,power_law_1.2,0.28682239055633546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,1,power_law_1.2,0.304422402381897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,1,power_law_1.2,0.3070208072662354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,1,power_law_1.2,0.34279038906097414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,1,power_law_1.2,0.3884351968765259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,1,power_law_1.2,0.4271359920501709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,1,power_law_1.2,0.513862419128418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,1,power_law_1.2,0.5993343830108643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,1,power_law_1.2,0.7675648212432862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,1,power_law_1.2,0.9263872146606446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,1,power_law_1.2,1.252780818939209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,1,power_law_1.2,1.5943936347961425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,1,power_law_1.2,2.250694465637207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,1,power_law_1.2,2.914950370788574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,1,power_law_1.2,3.589004898071289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,1,power_law_1.2,5.607603073120117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,1,power_law_1.2,10.9889404296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,1,power_law_1.2,0.09575039744377137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,1,power_law_1.2,0.10704640150070191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,1,power_law_1.2,0.11884160041809082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,1,power_law_1.2,0.15471999645233153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,1,power_law_1.2,0.20328960418701172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,1,power_law_1.2,0.23474559783935547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,1,power_law_1.2,0.32276480197906493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,1,power_law_1.2,0.33678081035614016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,1,power_law_1.2,0.3360447883605957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,1,power_law_1.2,0.34609920978546144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,1,power_law_1.2,0.3584000110626221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,1,power_law_1.2,0.369267201423645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,1,power_law_1.2,0.3890687942504883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,1,power_law_1.2,0.3996351957321167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,1,power_law_1.2,0.40407681465148926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,1,power_law_1.2,0.42641921043395997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,1,power_law_1.2,0.4432703971862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,1,power_law_1.2,0.5038591861724854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,1,power_law_1.2,0.021267199516296388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.2,0.5595647811889648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,16,balanced,0.03374933451414108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,1,power_law_1.2,0.022655999660491942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.2,0.6506432056427002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,16,balanced,0.035088000198205314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,1,power_law_1.2,0.02593280076980591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.2,0.7350336074829101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,16,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,1,power_law_1.2,0.03134720027446747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,16,balanced,0.0421066681543986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.2,0.9431424140930176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,16,balanced,0.04515199859937032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,1,power_law_1.2,0.04442239999771118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.2,1.1316736221313477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,16,balanced,0.047370667258898415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,1,power_law_1.2,0.06672000288963317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,16,balanced,0.0483893354733785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.2,1.5248191833496094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,16,balanced,0.04897599915663401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,1,power_law_1.2,0.07826560139656066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.2,1.9078271865844727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,16,balanced,0.04926399886608124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,1,power_law_1.2,0.08609920144081115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,16,balanced,0.04829333225886027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.2,2.681363105773926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,1,power_law_1.2,0.08833280205726624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,16,balanced,0.04660800099372864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.2,3.4490432739257812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,16,balanced,0.05392000079154968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,1,power_law_1.2,0.09107199907302857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,16,balanced,0.05023466547330221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.2,4.211999893188477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,1,power_law_1.2,0.0998207986354828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,16,balanced,0.050986667474110924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,1,power_law_1.2,0.10634880065917969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,16,balanced,0.0503359983364741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,1,power_law_1.2,6.51357421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,1,power_law_1.2,0.11458560228347778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,16,balanced,0.05243733525276184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,1,power_law_1.2,0.12569600343704224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,16,balanced,0.05031466484069824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,1,power_law_1.2,12.710758209228516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,1,power_law_1.2,0.11948800086975098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,16,balanced,0.05222400029500326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,1,power_law_1.2,0.12549760341644287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,16,balanced,0.050479998191197716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,1,power_law_1.2,0.13992960453033448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,1,power_law_1.2,0.14181120395660402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,16,balanced,0.054805333415667214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.2,0.15827200412750245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,16,balanced,0.05786666770776113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.2,0.19022719860076903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.2,0.21790080070495604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.2,0.32445440292358396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,16,balanced,0.07658666869004567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.2,0.37731199264526366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.2,0.4172351837158203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.2,0.4994815826416016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,16,balanced,0.0827946662902832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.2,0.6643392086029053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.2,0.8263296127319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.2,0.9790143966674805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,16,balanced,0.099589337905248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,1,power_law_1.2,1.4911423683166505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,16,balanced,0.03381866713364919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,16,balanced,0.034474665919939675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,1,power_law_1.2,2.8244607925415037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,16,balanced,0.03477333237727483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,16,balanced,0.03856533269087473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,16,balanced,0.11801600456237793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,16,balanced,0.04650666813055674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,16,balanced,0.04463466505209605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,16,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,16,balanced,0.04423466821511587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,16,balanced,0.044778664906819664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,16,balanced,0.0447573314110438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,16,balanced,0.04418666660785675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,16,balanced,0.042378668983777366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,16,balanced,0.04651733239491781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,16,balanced,0.1504906713962555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,16,balanced,0.04554133117198944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,16,balanced,0.04734933376312256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,16,balanced,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,16,balanced,0.04822933177153269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,16,balanced,0.05157866577307383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,16,balanced,0.05272533496220907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,16,balanced,0.05713599920272827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,16,balanced,0.18678400913874307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,16,balanced,0.060778667529424034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,16,balanced,0.08097066481908162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,16,balanced,0.09272533655166626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,16,balanced,0.22408533096313477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,16,balanced,0.12377599875132243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,16,balanced,0.1490559975306193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,16,balanced,0.20044267177581787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,16,balanced,0.33218133449554443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,16,balanced,0.2535039981206258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,16,balanced,0.03376533339420954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,16,balanced,0.03499199946721395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,16,balanced,0.036506667733192444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,16,balanced,0.3059999942779541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,16,balanced,0.040864000717798867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,16,balanced,0.045754666129748024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,16,balanced,0.04650666813055674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,16,balanced,0.04953599969546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,16,balanced,0.04747733473777771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,16,balanced,0.047050664822260536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,16,balanced,0.04716266691684723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,16,balanced,0.046767999728520714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,16,balanced,0.04890666902065277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,16,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,16,balanced,0.04882133503754934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,16,balanced,0.053077335158983864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,16,balanced,0.05287999908129374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,16,balanced,0.052842666705449425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,16,balanced,0.05451733370621999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,16,balanced,0.05840533475081126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,16,balanced,0.6150613228480021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,16,balanced,0.05906666815280914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,16,balanced,0.46647465229034424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,16,balanced,0.01964266722400983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,16,balanced,0.06483733157316844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,16,balanced,0.02011200040578842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,16,balanced,0.020282667130231857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,16,balanced,0.021695998807748158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,16,balanced,0.07704533139864604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,16,balanced,0.02383466561635335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,16,balanced,0.024613333245118458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,16,balanced,0.027994667490323383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,16,balanced,0.026170666019121807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,16,balanced,0.08501866459846497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,16,balanced,0.03809600075085958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,16,balanced,0.0367253323396047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,16,balanced,0.036288000643253326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,16,balanced,0.040074666341145836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,16,balanced,0.03880000114440918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,16,balanced,0.10949333508809407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,16,balanced,0.03622400015592575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,16,balanced,0.04631466666857401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,16,balanced,0.04477333525816599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,16,balanced,0.040421334405740104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,16,balanced,0.13172800342241922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,16,balanced,0.048911998669306435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,16,balanced,0.049141332507133484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,16,balanced,0.058090666929880776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,16,balanced,0.06469333171844482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,16,balanced,0.17466666301091513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,16,balanced,0.08130133152008057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,16,balanced,0.09787733356157939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,16,balanced,0.9026347001393636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,16,balanced,0.13177067041397095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,16,balanced,0.21313599745432535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,16,balanced,0.16385066509246826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,16,balanced,0.2591200073560079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,16,balanced,0.23161600033442178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,16,balanced,0.3002400000890096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,16,balanced,0.3798346519470215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,16,balanced,0.3675040006637573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,16,power_law_1.01,0.04296320080757141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,16,power_law_1.01,0.036595198512077334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,16,power_law_1.01,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,16,power_law_1.01,0.03898879885673523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,16,power_law_1.01,0.04164479970932007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,16,power_law_1.01,0.0462336003780365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,16,power_law_1.01,0.04679040014743805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,16,power_law_1.01,0.045407998561859134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,16,power_law_1.01,0.04612480103969574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,16,power_law_1.01,0.04648320078849792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,16,power_law_1.01,0.04800640046596527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,16,power_law_1.01,0.04801279902458191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,16,power_law_1.01,0.04847359955310822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,16,power_law_1.01,0.04913919866085052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,16,power_law_1.01,0.052243202924728394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,16,power_law_1.01,0.05058559775352478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,16,power_law_1.01,0.05359359979629517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,16,power_law_1.01,0.05780479907989502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,16,power_law_1.01,0.06762880086898804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,16,balanced,0.5725440184275309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,16,power_law_1.01,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,16,power_law_1.01,0.07854719758033753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,16,power_law_1.01,0.09745280146598816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,16,power_law_1.01,0.11534719467163086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,16,balanced,0.714629332224528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,16,power_law_1.01,0.14840960502624512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,16,power_law_1.01,0.19159040451049805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,16,power_law_1.01,0.242956805229187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,16,power_law_1.01,0.3071039915084839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,16,power_law_1.01,0.40940160751342775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,16,power_law_1.01,0.6173632144927979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,16,power_law_1.01,1.193887996673584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,16,power_law_1.01,0.04343039989471435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,16,power_law_1.01,0.03464959859848023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,16,power_law_1.01,0.03772799968719483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,16,power_law_1.01,0.03827199935913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,16,power_law_1.01,0.03893119990825653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,16,power_law_1.01,0.04092159867286682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,16,power_law_1.01,0.043372800946235655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,16,power_law_1.01,0.04268800020217896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,16,power_law_1.01,0.04249599874019623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,16,power_law_1.01,0.04312320053577423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,16,power_law_1.01,0.04387840032577515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,16,power_law_1.01,0.04340479969978332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,16,power_law_1.01,0.043321600556373595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,16,power_law_1.01,0.04506880044937134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,16,power_law_1.01,0.04816640019416809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,16,power_law_1.01,0.04984959959983826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,16,power_law_1.01,0.05120000243186951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,16,power_law_1.01,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,16,power_law_1.01,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,16,power_law_1.01,0.07100800275802613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,16,power_law_1.01,0.08379520177841186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,16,power_law_1.01,0.1033471941947937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,16,power_law_1.01,0.12620160579681397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,16,power_law_1.01,0.1831936001777649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,16,power_law_1.01,0.21850879192352296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,16,power_law_1.01,0.3114943981170654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,16,balanced,1.121999979019165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,16,power_law_1.01,0.4053952217102051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,16,power_law_1.01,0.5228799819946289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,16,power_law_1.01,0.7295040130615235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,16,power_law_1.01,1.5712767601013184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,16,power_law_1.01,0.043756800889968875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,16,power_law_1.01,0.03720319867134094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,16,power_law_1.01,0.03552640080451965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,16,power_law_1.01,0.039955198764801025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,16,power_law_1.01,0.041555199027061465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,16,power_law_1.01,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,16,power_law_1.01,0.04601599872112274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,16,power_law_1.01,0.04589439928531647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,16,power_law_1.01,0.04614399969577789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,16,power_law_1.01,0.04593920111656189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,16,power_law_1.01,0.046438398957252505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,16,power_law_1.01,0.018534399569034576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,16,power_law_1.01,0.047040000557899475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,16,power_law_1.01,0.01812479943037033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,16,power_law_1.01,0.047635200619697574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,16,power_law_1.01,0.01797119975090027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,16,power_law_1.01,0.048172798752784726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,16,power_law_1.01,0.01903360038995743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,16,power_law_1.01,0.020844799280166627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,16,power_law_1.01,0.022457599639892578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,16,power_law_1.01,0.025811201333999632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,16,power_law_1.01,0.025203201174736022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,16,power_law_1.01,0.036473599076271054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,16,power_law_1.01,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,16,power_law_1.01,0.03591040074825287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,16,power_law_1.01,0.039743998646736146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,16,power_law_1.01,0.037241598963737486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,16,power_law_1.01,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,16,power_law_1.01,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,16,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,16,power_law_1.01,0.035545599460601804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,16,power_law_1.01,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,16,power_law_1.01,0.043673598766326906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,16,power_law_1.01,0.04742400050163269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,16,power_law_1.01,0.05262719988822937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,16,power_law_1.01,0.06495360136032105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,16,power_law_1.01,0.07559679746627808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,16,power_law_1.01,0.10555520057678222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,16,power_law_1.01,0.12264959812164307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,16,power_law_1.01,0.17414400577545167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,16,power_law_1.01,0.2241919994354248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,16,power_law_1.01,0.24700798988342285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,16,power_law_1.01,0.4442111968994141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,16,power_law_1.01,0.8543999671936036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,16,power_law_1.2,0.04168959856033325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,16,power_law_1.2,0.03460479974746704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,16,power_law_1.2,0.037478399276733396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,16,power_law_1.2,0.03937920033931732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,16,power_law_1.2,0.0411327987909317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,16,power_law_1.2,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,16,power_law_1.2,0.047942399978637695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,16,power_law_1.2,0.040191999077796935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,16,power_law_1.2,0.05028480291366577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,16,power_law_1.2,0.03381119966506958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,16,power_law_1.2,0.049395200610160825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,16,power_law_1.2,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,16,power_law_1.2,0.04910080134868622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,16,power_law_1.2,0.03720960021018982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,16,power_law_1.2,0.04792320132255554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,16,power_law_1.2,0.037510401010513304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,16,power_law_1.2,0.05113599896430969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,16,power_law_1.2,0.04226559996604919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,16,power_law_1.2,0.050406402349472045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,16,power_law_1.2,0.043136000633239746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,16,power_law_1.2,0.04997760057449341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,16,power_law_1.2,0.042796799540519716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,16,power_law_1.2,0.05448960065841675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,16,power_law_1.2,0.04268159866333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,16,power_law_1.2,0.056569600105285646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,16,power_law_1.2,0.043315199017524716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,16,power_law_1.2,0.05960959792137146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,16,power_law_1.2,0.044940799474716187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,16,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,16,power_law_1.2,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,16,power_law_1.2,0.044710400700569156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,16,power_law_1.2,0.07409279942512512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,16,power_law_1.2,0.04593920111656189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,16,power_law_1.2,0.08164479732513427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,16,power_law_1.2,0.05074560046195984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,16,power_law_1.2,0.09395840167999267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,16,power_law_1.2,0.05367680191993714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,16,power_law_1.2,0.10370559692382812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,16,power_law_1.2,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,16,power_law_1.2,0.13105920553207398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,16,power_law_1.2,0.05931519865989685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,16,power_law_1.2,0.16151039600372313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,16,power_law_1.2,0.06538239717483521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,16,power_law_1.2,0.24360959529876708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,16,power_law_1.2,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,16,power_law_1.2,0.2986112117767334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,16,power_law_1.2,0.08685439825057983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,16,power_law_1.2,0.45162239074707033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,16,power_law_1.2,0.10629119873046874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,16,power_law_1.2,0.5936768054962158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,16,power_law_1.2,0.1350208044052124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,16,power_law_1.2,0.8407487869262695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,16,power_law_1.2,0.18922239542007446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,16,power_law_1.2,0.26800639629364015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,16,power_law_1.2,1.7525632858276368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,16,power_law_1.2,0.3603071928024292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,16,power_law_1.2,0.5095232009887696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,16,power_law_1.2,0.7306687831878662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,16,power_law_1.2,1.0004672050476073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,16,power_law_1.2,2.2795711517333985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,16,power_law_1.01,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,16,power_law_1.01,0.05280640125274658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,16,power_law_1.01,0.05489919781684875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,16,power_law_1.01,0.05902079939842224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,16,power_law_1.01,0.06433280110359192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,16,power_law_1.01,0.06993280053138733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,16,power_law_1.01,0.08347520232200623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,16,power_law_1.01,0.10097279548645019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,16,power_law_1.01,0.12387839555740357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,16,power_law_1.01,0.15133440494537354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,16,power_law_1.2,0.046003198623657225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,16,power_law_1.01,0.19653120040893554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,16,power_law_1.2,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,16,power_law_1.01,0.2638015985488892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,16,power_law_1.2,0.036582401394844054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,16,power_law_1.01,0.3338304042816162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,16,power_law_1.2,0.038489601016044615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,16,power_law_1.2,0.04136959910392761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,16,power_law_1.01,0.4672959804534912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,16,power_law_1.2,0.0456959992647171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,16,power_law_1.01,0.6586880207061767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,16,power_law_1.2,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,16,power_law_1.01,1.409222412109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,16,power_law_1.2,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,16,power_law_1.2,0.046515199542045596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,16,power_law_1.2,0.04599680006504059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,16,power_law_1.2,0.047276800870895384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,16,power_law_1.2,0.04924800097942352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,16,power_law_1.2,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,16,power_law_1.2,0.04917759895324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,16,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,16,power_law_1.2,0.05448319911956787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,16,power_law_1.2,0.05681279897689819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,16,power_law_1.2,0.06298239827156067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,16,power_law_1.2,0.06766080260276794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,16,power_law_1.2,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,16,power_law_1.2,0.08688639998435974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,16,power_law_1.2,0.10867840051651001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,16,power_law_1.2,0.13233920335769653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,16,power_law_1.2,0.17295360565185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,16,power_law_1.2,0.21811199188232422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,16,power_law_1.2,0.3652672052383423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,16,power_law_1.2,0.40517120361328124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,16,power_law_1.2,0.6070400238037109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,16,power_law_1.2,0.01865600049495697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,16,power_law_1.2,0.017497600615024568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,16,power_law_1.2,1.0273471832275392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,16,power_law_1.2,0.01852799952030182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,16,power_law_1.2,1.8108736038208009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,16,power_law_1.2,0.019705599546432494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,8,balanced,0.032586666444937386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,16,power_law_1.2,0.021465599536895752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,8,balanced,0.03626666714747747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,16,power_law_1.2,0.023097600042819976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,8,balanced,0.037632000943024956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,16,power_law_1.2,0.02678399980068207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,8,balanced,0.04228800038496653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,8,balanced,0.04909333089987437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,16,power_law_1.2,0.025567999482154845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,8,balanced,0.049141332507133484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,16,power_law_1.2,0.03692800104618073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,8,balanced,0.04965866605440775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,16,power_law_1.2,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,8,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,16,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,8,balanced,0.04901333153247833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,16,power_law_1.2,0.038867199420928956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,8,balanced,0.049045334259668984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,8,balanced,0.04781866570313772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,16,power_law_1.2,0.039929598569869995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,8,balanced,0.048800001541773476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,16,power_law_1.2,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,8,balanced,0.047007997830708824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,16,power_law_1.2,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,8,balanced,0.048437332113583885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,16,power_law_1.2,0.035897600650787356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,8,balanced,0.05315199991067251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,16,power_law_1.2,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,8,balanced,0.051327998439470925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,16,power_law_1.2,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,16,power_law_1.2,0.04289279878139496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,8,balanced,0.05263466636339823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,16,power_law_1.2,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,8,balanced,0.053317333261171974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,16,power_law_1.2,0.05859839916229248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,8,balanced,0.05745066702365875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,16,power_law_1.2,0.06503040194511414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,16,power_law_1.2,0.07985280156135559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,8,balanced,0.060602664947509766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,16,power_law_1.2,0.10943360328674316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,16,power_law_1.2,0.1347584009170532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,8,balanced,0.06683733562628429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,16,power_law_1.2,0.19725439548492432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,16,power_law_1.2,0.2575808048248291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,16,power_law_1.2,0.3010175943374634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,8,balanced,0.08893866340319316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,16,power_law_1.2,0.4706560134887695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,16,power_law_1.2,0.9295680046081543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,8,balanced,0.09470400214195251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,8,balanced,0.11629866560300191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,8,balanced,0.13778666655222574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,8,balanced,0.18583466609319052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,8,balanced,0.2380853295326233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,8,balanced,0.0315733328461647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,8,balanced,0.03474666674931844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,8,balanced,0.0359253336985906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,8,balanced,0.04043733328580856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,8,balanced,0.04341333111127218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,8,balanced,0.04475200176239014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,8,balanced,0.04680533210436503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,8,balanced,0.045040001471837364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,8,balanced,0.04462933540344238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,8,balanced,0.04479999840259552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,8,balanced,0.29016000032424927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,8,balanced,0.045456002155939736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,8,balanced,0.04801600178082784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,8,balanced,0.046816001335779824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,8,balanced,0.04710933566093445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,8,balanced,0.05227200190226237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,8,balanced,0.050842667619387306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,8,balanced,0.05085866649945577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,8,balanced,0.054469332098960876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,8,balanced,0.05857066810131073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,8,balanced,0.06341866652170818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,8,balanced,0.06949866811434428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,8,balanced,0.03385599950949351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,8,balanced,0.03658666710058848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,8,balanced,0.03869866579771042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,8,balanced,0.09707199533780415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,8,balanced,0.04433066646258036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,8,balanced,0.05286400020122528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,8,balanced,0.434442679087321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,8,balanced,0.05287466446558634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,8,balanced,0.05087999999523163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,8,balanced,0.051221330960591636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,8,balanced,0.11417067050933838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,8,balanced,0.05259733398755392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,8,balanced,0.05093333125114441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,8,balanced,0.0507893313964208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,8,balanced,0.05093333125114441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,8,balanced,0.05295466880003611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,8,balanced,0.1469013293584188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,8,balanced,0.0526506652434667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,8,balanced,0.055018668373425804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,8,balanced,0.05754666527112325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,8,balanced,0.05925333499908447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,8,balanced,0.06332266827424367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,8,balanced,0.1792479952176412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,8,balanced,0.0643039991458257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,8,balanced,0.07102400064468384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,8,balanced,0.07577066620190938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,8,balanced,0.0979146659374237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,8,balanced,0.24953599770863852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,8,balanced,0.10937600334485371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,8,balanced,0.019248000035683315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,8,balanced,0.020010666300853092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,8,balanced,0.14854400356610617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,8,balanced,0.020618667205174763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,8,balanced,0.022463999688625336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,8,balanced,0.022602667411168415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,8,balanced,0.023797333240509033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,8,balanced,0.32067734003067017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,8,balanced,0.026522666215896606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,8,balanced,0.025920001169045765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,8,balanced,0.030602666238943737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,8,balanced,0.030095999439557392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,8,balanced,0.18001067638397217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,8,balanced,0.8111200332641602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,8,balanced,0.028981332977612812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,8,balanced,0.04055999964475632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,8,balanced,0.04070399949947993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,8,balanced,0.04055466751257578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,8,balanced,0.04586133360862732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,8,balanced,0.04642133414745331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,8,balanced,0.04418133199214935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,8,balanced,0.048810665806134544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,8,balanced,0.3965280055999756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,8,balanced,0.24773865938186646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,8,balanced,0.052149335543314614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,8,balanced,0.05630933245023092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,8,balanced,0.0603359987338384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,8,balanced,0.07367466886838277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,8,balanced,0.31510400772094727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,8,balanced,0.08703999718030293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,8,balanced,0.11668266852696736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,8,balanced,0.6141226689020792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,8,balanced,0.14086400469144186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,8,balanced,0.38075733184814453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,8,balanced,0.19575466712315878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,8,balanced,0.2507839997609456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,8,balanced,0.5796213150024414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,8,balanced,0.3032853404680888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,8,balanced,1.1876586278279622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,8,balanced,0.47042667865753174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,8,power_law_1.01,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,8,power_law_1.01,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,8,power_law_1.01,0.036006399989128114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,8,power_law_1.01,0.03898879885673523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,8,power_law_1.01,0.040755200386047366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,8,power_law_1.01,0.04791040122509003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,8,power_law_1.01,0.046623998880386354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,8,power_law_1.01,0.04760960042476654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,8,balanced,1.098527987798055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,8,power_law_1.01,0.04785920083522797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,8,power_law_1.01,0.04863359928131104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,8,power_law_1.01,0.04903680086135864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,8,power_law_1.01,0.049702399969100954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,8,power_law_1.01,0.04930559992790222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,8,power_law_1.01,0.04927360117435455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,8,power_law_1.01,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,8,power_law_1.01,0.05502079725265503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,8,power_law_1.01,0.05759360194206238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,8,power_law_1.01,0.05980799794197082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,8,power_law_1.01,0.0667519986629486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,8,power_law_1.01,0.07923840284347534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,8,power_law_1.01,0.08521599769592285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,8,power_law_1.01,0.1028223991394043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,8,power_law_1.01,0.12092800140380859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,8,power_law_1.01,0.16272640228271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,8,power_law_1.01,0.19267840385437013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,8,power_law_1.01,0.2682431936264038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,8,power_law_1.01,0.30819199085235593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,8,power_law_1.01,0.4108863830566406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,8,power_law_1.01,0.6234367847442627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,8,power_law_1.01,1.243455982208252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,8,balanced,0.9140426317850748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,8,power_law_1.01,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,8,power_law_1.01,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,8,power_law_1.01,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,8,power_law_1.01,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,8,power_law_1.01,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,8,power_law_1.01,0.04353919923305512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,8,power_law_1.01,0.04328320026397705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,8,power_law_1.01,0.04312959909439087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,8,power_law_1.01,0.043161600828170776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,8,power_law_1.01,0.044326400756835936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,8,power_law_1.01,0.04511359930038452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,8,power_law_1.01,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,8,power_law_1.01,0.046937599778175354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,8,power_law_1.01,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,8,power_law_1.01,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,8,power_law_1.01,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,8,power_law_1.01,0.054739201068878175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,8,power_law_1.01,0.05798400044441223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,8,power_law_1.01,0.06359040141105651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,8,power_law_1.01,0.07437440156936645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,8,power_law_1.01,0.09089279770851136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,8,power_law_1.01,0.11929600238800049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,8,power_law_1.01,0.1437440037727356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,8,power_law_1.01,0.18931200504302978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,8,power_law_1.01,0.24427518844604493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,8,power_law_1.01,0.3012991905212402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,8,power_law_1.01,0.501523208618164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,8,power_law_1.01,0.5862592220306396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,8,power_law_1.01,0.9283712387084961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,8,power_law_1.01,0.04252800047397613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,8,power_law_1.01,1.6024703979492188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,8,power_law_1.01,0.03803519904613495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,8,power_law_1.01,0.038412800431251524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,8,power_law_1.01,0.042694398760795595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,8,power_law_1.01,0.04482559859752655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,8,power_law_1.01,0.049983999133110045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,8,power_law_1.01,0.05022720098495483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,8,power_law_1.01,0.05013759732246399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,8,power_law_1.01,0.050387197732925416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,8,power_law_1.01,0.05021439790725708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,8,power_law_1.01,0.05125759840011597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,8,power_law_1.01,0.05162879824638367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,8,power_law_1.01,0.05256320238113403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,8,power_law_1.01,0.05312640070915222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,8,power_law_1.01,0.0562175989151001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,8,power_law_1.01,0.05735679864883423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,8,power_law_1.01,0.06172159910202026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,8,power_law_1.01,0.06564480066299438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,8,power_law_1.01,0.07475200295448303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,8,power_law_1.2,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,8,power_law_1.01,0.08429440259933471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,8,power_law_1.2,0.035155200958251955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,8,power_law_1.01,0.09710080027580262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,8,power_law_1.2,0.03683840036392212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,8,power_law_1.01,0.12360960245132446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,8,power_law_1.01,0.01785600036382675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,8,power_law_1.2,0.03877759873867035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,8,power_law_1.01,0.14092799425125122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,8,power_law_1.01,0.018681600689888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,8,power_law_1.2,0.040659201145172116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,8,power_law_1.01,0.19837440252304078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,8,power_law_1.01,0.0194815993309021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,8,power_law_1.2,0.045587199926376346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,8,power_law_1.01,0.018668800592422485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,8,power_law_1.01,0.22200961112976075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,8,power_law_1.2,0.04838399887084961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,8,power_law_1.01,0.020236800611019134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,8,power_law_1.01,0.3797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,8,power_law_1.2,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,8,power_law_1.01,0.02197760045528412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,8,power_law_1.01,0.44907522201538086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,8,power_law_1.2,0.04872959852218628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,8,power_law_1.01,0.023372800648212434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,8,power_law_1.01,0.5954815864562988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,8,power_law_1.2,0.04833920001983642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,8,power_law_1.01,0.02336000055074692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,8,power_law_1.01,0.8167424201965332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,8,power_law_1.2,0.04922240078449249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,8,power_law_1.01,0.026144000887870788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,8,power_law_1.2,0.0491456001996994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,8,power_law_1.01,1.7555839538574218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,8,power_law_1.01,0.02650879919528961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,8,power_law_1.2,0.0472896009683609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,8,power_law_1.01,0.026982399821281432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,8,power_law_1.2,0.05051519870758057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,8,power_law_1.01,0.03800320029258728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,8,power_law_1.2,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,8,power_law_1.01,0.036499199271202085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,8,power_law_1.2,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,8,power_law_1.01,0.03734399974346161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,8,power_law_1.2,0.058233600854873654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,8,power_law_1.01,0.04132480025291443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,8,power_law_1.2,0.06099839806556702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,8,power_law_1.01,0.038380798697471616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,8,power_law_1.2,0.0716159999370575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,8,power_law_1.01,0.04069760143756866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,8,power_law_1.2,0.0794368028640747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,8,power_law_1.01,0.041247999668121337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,8,power_law_1.2,0.0874176025390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,8,power_law_1.01,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,8,power_law_1.2,0.10440959930419921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,8,power_law_1.01,0.04903680086135864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,8,power_law_1.2,0.12595200538635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,8,power_law_1.01,0.051648002862930295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,8,power_law_1.01,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,8,power_law_1.2,0.18717440366744995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,8,power_law_1.01,0.07317759990692138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,8,power_law_1.2,0.21438078880310057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,8,power_law_1.01,0.09342079758644103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,8,power_law_1.2,0.3006783962249756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,8,power_law_1.01,0.11409920454025269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,8,power_law_1.2,0.43228797912597655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,8,power_law_1.01,0.1568127989768982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,8,power_law_1.2,0.5521024227142334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,8,power_law_1.01,0.2070080041885376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,8,power_law_1.2,0.840550422668457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,8,power_law_1.01,0.2484287977218628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,8,power_law_1.2,1.6878528594970703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,8,power_law_1.01,0.3584127902984619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,8,power_law_1.01,0.7501567840576172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,8,power_law_1.2,0.03668479919433594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,8,power_law_1.2,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,8,power_law_1.2,0.03463680148124695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,8,power_law_1.2,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,8,power_law_1.2,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,8,power_law_1.2,0.04284160137176514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,8,power_law_1.2,0.04335359930992126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,8,power_law_1.2,0.04336000084877014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,8,power_law_1.2,0.04437119960784912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,8,power_law_1.2,0.04501760005950928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,8,power_law_1.2,0.04477440118789673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,8,power_law_1.2,0.04620159864425659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,8,power_law_1.2,0.04739840030670166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,8,power_law_1.2,0.048127999901771544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,8,power_law_1.2,0.0517632007598877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,8,power_law_1.2,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,8,power_law_1.2,0.054451197385787964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,8,power_law_1.2,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,8,power_law_1.2,0.0656000018119812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,8,power_law_1.2,0.08174719810485839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,8,power_law_1.2,0.0938368022441864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,8,power_law_1.2,0.1298815965652466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,8,power_law_1.2,0.15362559556961058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,8,power_law_1.2,0.24224638938903809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,8,power_law_1.2,0.2599040031433105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,8,power_law_1.2,0.4227712154388428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,8,power_law_1.2,0.5134592056274414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,8,power_law_1.2,0.8573247909545898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,8,power_law_1.2,1.1791935920715333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,8,power_law_1.2,2.2602943420410155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,8,power_law_1.2,0.04287360012531281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,8,power_law_1.2,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,8,power_law_1.2,0.036735999584198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,8,power_law_1.2,0.0416128009557724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,8,power_law_1.2,0.04307839870452881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,8,power_law_1.2,0.04865280091762543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,8,power_law_1.2,0.049728000164031984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,8,power_law_1.2,0.049856001138687135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,8,power_law_1.2,0.05089920163154602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,8,power_law_1.2,0.05056639909744263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,8,power_law_1.2,0.05045120120048523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,8,power_law_1.2,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,8,power_law_1.2,0.05187199711799621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,8,power_law_1.2,0.0538752019405365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,8,power_law_1.2,0.05803520083427429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,8,power_law_1.2,0.05980160236358643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,8,power_law_1.2,0.06220800280570984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,8,power_law_1.2,0.01701119989156723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,8,power_law_1.2,0.06881279945373535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,8,power_law_1.2,0.017612800002098083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,8,power_law_1.2,0.07507200241088867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,8,power_law_1.2,0.018380799889564516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,4,balanced,0.033973333736260734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,8,power_law_1.2,0.08876799941062927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,8,power_law_1.2,0.017638400197029114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,4,balanced,0.03631466627120972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,8,power_law_1.2,0.10725120306015015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,8,power_law_1.2,0.019494399428367615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,4,balanced,0.038560000558694206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,8,power_law_1.2,0.02157440036535263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,8,power_law_1.2,0.13600640296936034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,4,balanced,0.04301866888999939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,8,power_law_1.2,0.02215680032968521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,8,power_law_1.2,0.1587007999420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,4,balanced,0.05095999936262766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,4,balanced,0.05287466446558634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,8,power_law_1.2,0.02248319983482361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,8,power_law_1.2,0.21688320636749267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,4,balanced,0.052671998739242554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,8,power_law_1.2,0.025625601410865784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,8,power_law_1.2,0.27693440914154055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,4,balanced,0.052890668312708534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,8,power_law_1.2,0.026054400205612182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,4,balanced,0.04951466619968414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,8,power_law_1.2,0.44320001602172854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,8,power_law_1.2,0.024902400374412537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,4,balanced,0.0528053343296051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,8,power_law_1.2,0.547603178024292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,8,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,4,balanced,0.04978133241335551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,8,power_law_1.2,0.6590015888214111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,8,power_law_1.2,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,4,balanced,0.05237866441408793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,8,power_law_1.2,0.036652800440788266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,8,power_law_1.2,1.1261119842529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,4,balanced,0.05262400209903717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,8,power_law_1.2,0.04015359878540039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,4,balanced,0.051141331593195595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,8,power_law_1.2,2.4138240814208984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,8,power_law_1.2,0.03946239948272705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,4,balanced,0.056426664193471275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,8,power_law_1.2,0.03989759981632233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,4,balanced,0.05648000041643778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,8,power_law_1.2,0.039289599657058714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,4,balanced,0.0569706658522288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,8,power_law_1.2,0.040403199195861814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,4,balanced,0.06123200058937073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,8,power_law_1.2,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,8,power_law_1.2,0.05225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,4,balanced,0.06477866570154826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,8,power_law_1.2,0.06430720090866089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,8,power_law_1.2,0.0759935975074768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,4,balanced,0.07446933289368947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,8,power_law_1.2,0.09912959933280945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,8,power_law_1.2,0.1187391996383667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,4,balanced,0.08226133386294048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,8,power_law_1.2,0.1627776026725769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,8,power_law_1.2,0.21587200164794923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,8,power_law_1.2,0.2591808080673218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,4,balanced,0.10626133282979329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,8,power_law_1.2,0.39612159729003904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,8,power_law_1.2,0.811520004272461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,4,balanced,0.11814933021863301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,4,balanced,0.15026666720708212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,4,balanced,0.19090133905410767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,4,balanced,0.2674720088640849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,4,balanced,0.3452479839324951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,4,balanced,0.03387200087308884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,4,balanced,0.034917332231998444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,4,balanced,0.03697066754102707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,4,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,4,balanced,0.04770666857560476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,4,balanced,0.04834666848182678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,4,balanced,0.05026666820049286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,4,balanced,0.04877333343029022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,4,balanced,0.04867733518282572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,4,balanced,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,4,balanced,0.4198559919993083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,4,balanced,0.04877866804599762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,4,balanced,0.052383999029795326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,4,balanced,0.050586665670077004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,4,balanced,0.052613332867622375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,4,balanced,0.05726933479309082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,4,balanced,0.0572320024172465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,4,balanced,0.058730666836102806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,4,balanced,0.0645653357108434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,4,balanced,0.06771733363469441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,4,balanced,0.07503466804822286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,4,balanced,0.09125333031018575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,4,balanced,0.12652267018953958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,4,balanced,0.6331520080566406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,4,power_law_1.01,0.03651840090751648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,4,balanced,0.14849600195884705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,4,power_law_1.01,0.0349375993013382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,4,power_law_1.01,0.03472639918327332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,4,power_law_1.01,0.03975679874420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,4,power_law_1.01,0.039878401160240176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,4,power_law_1.01,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,4,balanced,0.19775466124216715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,4,power_law_1.01,0.04747520089149475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,4,power_law_1.01,0.04930559992790222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,4,power_law_1.01,0.04859519898891449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,4,power_law_1.01,0.048851200938224794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,4,power_law_1.01,0.04991360008716583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,4,power_law_1.01,0.05112959742546082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,4,power_law_1.01,0.051500797271728516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,4,power_law_1.01,0.05251200199127197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,4,balanced,0.24671467145284018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,4,power_law_1.01,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,4,power_law_1.01,0.05678079724311828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,4,power_law_1.01,0.05928320288658142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,4,power_law_1.01,0.06669440269470214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,4,power_law_1.01,0.07109119892120361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,4,power_law_1.01,0.08798720240592957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,4,power_law_1.01,0.10050560235977173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,4,power_law_1.01,0.11684479713439941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,4,power_law_1.01,0.13528319597244262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,4,balanced,0.3513866662979126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,4,power_law_1.01,0.19221760034561158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,4,power_law_1.01,0.03578880131244659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,4,power_law_1.01,0.23134078979492187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,4,power_law_1.01,0.03333120048046112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,4,power_law_1.01,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,4,power_law_1.01,0.32316160202026367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,4,power_law_1.01,0.03866879940032959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,4,power_law_1.01,0.45980157852172854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,4,power_law_1.01,0.04062080085277557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,4,power_law_1.01,0.5263296127319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,4,power_law_1.01,0.04537599980831146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,4,power_law_1.01,0.8136639595031738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,4,power_law_1.01,0.04593920111656189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,4,power_law_1.01,1.5617919921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,4,power_law_1.01,0.047014400362968445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,4,power_law_1.01,0.047200000286102294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,4,power_law_1.01,0.04760960042476654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,4,power_law_1.01,0.05002880096435547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,4,power_law_1.01,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,4,balanced,0.47250131766001385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,4,power_law_1.01,0.05089920163154602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,4,power_law_1.01,0.052742397785186766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,4,power_law_1.01,0.057183998823165896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,4,power_law_1.01,0.05790719985961914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,4,balanced,1.1979786554972331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,4,power_law_1.01,0.06149759888648987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,4,power_law_1.01,0.06943359971046448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,4,power_law_1.01,0.07484800219535828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,4,power_law_1.01,0.09767040014266967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,4,power_law_1.01,0.1135103940963745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,4,power_law_1.01,0.14472960233688353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,4,power_law_1.01,0.17558399438858033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,4,power_law_1.01,0.22507519721984864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,4,power_law_1.01,0.2883903980255127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,4,power_law_1.01,0.44869117736816405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,4,balanced,0.583242654800415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,4,power_law_1.01,0.6042751789093017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,4,power_law_1.01,0.7796671867370606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,4,power_law_1.01,1.2618559837341308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,4,power_law_1.01,2.3569408416748048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,4,balanced,0.9105546474456787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,4,power_law_1.2,0.036160001158714296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,4,power_law_1.2,0.033529600501060484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,4,balanced,1.7797600428263347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,4,power_law_1.2,0.03664000034332275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,4,power_law_1.2,0.03969280123710632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,4,power_law_1.2,0.04092159867286682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,4,power_law_1.2,0.04833920001983642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,4,power_law_1.2,0.04702079892158508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,4,power_law_1.2,0.04747520089149475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,4,power_law_1.2,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,4,power_law_1.2,0.04521600008010864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,4,power_law_1.2,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,4,power_law_1.2,0.05050879716873169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,4,power_law_1.2,0.050400000810623166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,4,power_law_1.2,0.051072001457214355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,4,power_law_1.2,0.03542400002479553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,4,power_law_1.2,0.057081598043441775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,4,power_law_1.2,0.033580800890922545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,4,power_law_1.2,0.057158398628234866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,4,power_law_1.2,0.034483200311660765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,4,power_law_1.2,0.06224640011787415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,4,power_law_1.2,0.03874559998512268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,4,power_law_1.2,0.06750079989433289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,4,power_law_1.2,0.04041599929332733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,4,power_law_1.2,0.07900159955024719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,4,power_law_1.2,0.04545280039310455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,4,power_law_1.2,0.08609279990196228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,4,power_law_1.2,0.04643200039863586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,4,power_law_1.2,0.10119680166244507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,4,power_law_1.2,0.04844799935817719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,4,power_law_1.2,0.12229119539260865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,4,power_law_1.2,0.04704639911651611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,4,power_law_1.2,0.1548416018486023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,16,balanced,0.048298666874567665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,4,power_law_1.2,0.0493120014667511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,4,power_law_1.2,0.18313599824905397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,16,balanced,0.05107733110586802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,4,power_law_1.2,0.05001599788665771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,4,power_law_1.2,0.24639360904693602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,16,balanced,0.06144000093142191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,4,power_law_1.2,0.05137280225753784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,4,power_law_1.2,0.36138238906860354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,16,balanced,0.07559466858704884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,16,balanced,0.09752532839775085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,4,power_law_1.2,0.050297600030899045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,4,power_law_1.2,0.5059391975402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,16,balanced,0.10347732901573181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,4,power_law_1.2,0.053478401899337766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,4,power_law_1.2,0.5849152088165284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,16,balanced,0.10547199845314026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,4,power_law_1.2,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,16,balanced,0.10539733370145161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,4,power_law_1.2,0.9994879722595215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,4,power_law_1.2,0.05943040251731872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,16,balanced,0.11740799744923909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,4,power_law_1.2,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,4,power_law_1.2,1.8189888000488281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,16,balanced,0.11953066786130269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,4,power_law_1.2,0.07125120162963867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,16,balanced,0.10553600390752156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,4,power_law_1.2,0.07954559922218322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,16,balanced,0.11054933071136475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,16,balanced,0.1002346674601237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,4,power_law_1.2,0.09991679787635803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,16,balanced,0.10459199547767639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,4,power_law_1.2,0.11091840267181396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,4,power_law_1.2,0.1480191946029663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,16,balanced,0.1072106659412384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,4,power_law_1.2,0.1851263999938965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,16,balanced,0.1020960013071696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,4,power_law_1.2,0.25791358947753906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,16,balanced,0.10595200459162395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,4,power_law_1.2,0.29285120964050293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,16,balanced,0.10315733154614766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,4,power_law_1.2,0.4742015838623047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,16,balanced,0.10547733306884766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,4,power_law_1.2,0.5888959884643554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,4,power_law_1.2,0.889305591583252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,16,balanced,0.11012267072995503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,4,power_law_1.2,1.4903743743896485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,4,power_law_1.2,2.9598272323608397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,16,balanced,0.11639466881752014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,16,balanced,0.15059733390808105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,16,balanced,0.15104533235232034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,16,balanced,0.1908479928970337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,16,balanced,0.23161600033442178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,16,balanced,0.30736533800760907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,16,balanced,0.3935679992039998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,16,balanced,0.4792213439941406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,16,balanced,0.03996799886226654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,16,balanced,0.04068800061941147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,16,balanced,0.0430026650428772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,16,balanced,0.05072000126043955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,16,balanced,0.07492266595363617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,16,balanced,0.0765119989713033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,16,balanced,0.07346666852633159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,16,balanced,0.07317333420117696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,16,balanced,0.07679999868075053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,16,balanced,0.07474666833877563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,16,balanced,0.07364266614119212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,16,balanced,0.07540800174077351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,16,balanced,0.07683733105659485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,16,balanced,0.07533333202203114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,16,balanced,0.07946133116881053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,16,balanced,0.07977066437403361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,16,balanced,0.08411733309427898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,16,balanced,0.08566932876904805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,16,balanced,0.08807999889055888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,16,balanced,0.7263253529866537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,16,balanced,0.10048533479372661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,16,balanced,0.040965333580970764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,16,balanced,0.040821333726247154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,16,balanced,0.10920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,16,balanced,0.04496533175309499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,16,balanced,0.05952000121275584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,16,balanced,0.07190399865309398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,16,balanced,0.0742986649274826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,16,balanced,0.06494933366775513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,16,balanced,0.13879467050234476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,16,balanced,0.06761066615581512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,16,balanced,0.06956799825032552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,16,balanced,0.06695466736952464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,16,balanced,0.06509333352247874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,16,balanced,0.1536960005760193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,16,balanced,0.06987200180689494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,16,balanced,0.06925333539644878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,16,balanced,0.06926399966080983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,16,balanced,0.07153599957625072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,16,balanced,0.0777759999036789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,16,balanced,0.07659199833869934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,16,balanced,0.20486400524775186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,16,balanced,0.07699733475844066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,16,balanced,0.07973866661389668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,16,balanced,0.08509332935015361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,16,balanced,0.2512320081392924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,16,balanced,0.0911253293355306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,16,balanced,0.1223466694355011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,16,balanced,0.02362666775782903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,16,balanced,0.024911999702453613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,16,balanced,0.024490666886170704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,16,balanced,0.02442666639884313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,16,balanced,0.029968000948429108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,16,balanced,0.1313706636428833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,16,balanced,0.03429866582155228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,16,balanced,0.04794666667779287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,16,balanced,0.043525333205858864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,16,balanced,0.34990934530893963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,16,balanced,0.06386133531729381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,16,balanced,0.062314664324124656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,16,balanced,0.061205332477887474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,16,balanced,0.1730133295059204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,16,balanced,0.06750399867693584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,16,balanced,0.06435733536879222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,16,balanced,1.4068640073140461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,16,balanced,0.06043733159701029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,16,balanced,0.08063466846942902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,16,balanced,0.08025600016117096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,16,balanced,0.20898133516311646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,16,balanced,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,16,balanced,0.08057599763075511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,16,balanced,0.44725334644317627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,16,balanced,0.09494400024414062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,16,balanced,0.10771733522415161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,16,balanced,0.28410132726033527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,16,balanced,0.12524267037709555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,16,balanced,0.16034666697184244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,16,balanced,0.1921173334121704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,16,balanced,0.35797866185506183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,16,balanced,0.5469439824422201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,16,balanced,0.28380799293518066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,16,balanced,0.34681065877278644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,16,balanced,0.43166399002075195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,16,balanced,0.5039360125859579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,16,balanced,0.8374933401743571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,16,balanced,0.664901336034139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,16,balanced,0.6561973492304484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,16,balanced,0.8304426670074463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,16,balanced,1.314352035522461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,16,power_law_1.01,0.0948736011981964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,16,balanced,1.2564586798350017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,16,balanced,1.6228639284769695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,16,power_law_1.01,0.05928959846496582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,16,power_law_1.01,0.06344320178031922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,16,power_law_1.01,0.07411199808120728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,16,power_law_1.01,0.08301439881324768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,16,power_law_1.01,0.09910399913787842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,16,power_law_1.01,0.10055680274963379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,16,power_law_1.01,0.11153919696807861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,16,power_law_1.01,0.1019327998161316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,16,power_law_1.01,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,16,power_law_1.01,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,16,power_law_1.01,0.10537600517272949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,16,power_law_1.01,0.10214400291442871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,16,power_law_1.01,0.10337280035018921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,16,power_law_1.01,0.107750403881073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,16,power_law_1.01,0.11278719902038574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,16,power_law_1.01,0.11427839994430541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,16,power_law_1.01,0.12202240228652954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,16,power_law_1.01,0.139302396774292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,16,power_law_1.01,0.15921920537948608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,16,power_law_1.01,0.183404803276062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,16,power_law_1.01,0.21091198921203613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,16,power_law_1.01,0.25836160182952883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,16,power_law_1.01,0.35119359493255614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,16,power_law_1.01,0.41610240936279297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,16,power_law_1.01,0.5942656040191651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,16,power_law_1.01,0.8295295715332032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,16,power_law_1.01,1.0303423881530762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,16,power_law_1.01,1.7683904647827149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,16,balanced,2.5956106185913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,16,power_law_1.01,3.7756542205810546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,16,power_law_1.01,0.07548800110816956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,16,power_law_1.01,0.05356159806251526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,16,power_law_1.01,0.045772799849510194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,16,power_law_1.01,0.056934398412704465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,16,power_law_1.01,0.057529598474502563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,16,power_law_1.01,0.07340160012245178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,16,power_law_1.01,0.07292159795761108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,16,power_law_1.01,0.07366399765014649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,16,power_law_1.01,0.07245439887046815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,16,power_law_1.01,0.07354239821434021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,16,power_law_1.01,0.0763584017753601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,16,power_law_1.01,0.0775871992111206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,16,power_law_1.01,0.07425280213356018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,16,power_law_1.01,0.07365120053291321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,16,power_law_1.01,0.07688959836959838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,16,power_law_1.01,0.05767040252685547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,16,power_law_1.01,0.07987200021743775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,16,power_law_1.01,0.045158401131629944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,16,power_law_1.01,0.0821183979511261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,16,power_law_1.01,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,16,power_law_1.01,0.08834559917449951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,16,power_law_1.01,0.05953279733657837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,16,power_law_1.01,0.0969215989112854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,16,power_law_1.01,0.07227519750595093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,16,power_law_1.01,0.10762879848480225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,16,power_law_1.01,0.0699072003364563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,16,power_law_1.01,0.12580480575561523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,16,power_law_1.01,0.06595199704170226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,16,power_law_1.01,0.14382719993591309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,16,power_law_1.01,0.06718720197677612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,16,power_law_1.01,0.18744319677352905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,16,power_law_1.01,0.0693120002746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,16,power_law_1.01,0.22216320037841797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,16,power_law_1.01,0.07028480172157288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,16,power_law_1.01,0.3313407897949219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,16,power_law_1.01,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,16,power_law_1.01,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,16,power_law_1.01,0.4013120174407959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,16,power_law_1.01,0.07146880030632019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,16,power_law_1.01,0.5732351779937744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,16,power_law_1.01,0.07582079768180847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,16,power_law_1.01,0.8878591537475586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,16,power_law_1.01,0.08025599718093872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,16,power_law_1.01,0.8027327537536622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,16,power_law_1.01,0.08549119830131531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,16,power_law_1.01,1.4282367706298829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,16,power_law_1.01,0.09141759872436524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,16,power_law_1.01,2.6506111145019533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.01,0.10170880556106568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,16,power_law_1.2,0.10958720445632934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.01,0.11569279432296753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,16,power_law_1.2,0.05793920159339905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.01,0.1397312045097351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,16,power_law_1.01,0.026214399933815004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,16,power_law_1.2,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,16,power_law_1.01,0.025644800066947936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,16,power_law_1.2,0.0719488024711609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.01,0.1759935975074768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,16,power_law_1.01,0.02316800057888031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,16,power_law_1.2,0.07575680017471313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.01,0.2042304039001465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,16,power_law_1.01,0.023712000250816344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,16,power_law_1.2,0.1016319990158081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.01,0.2852479934692383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,16,power_law_1.01,0.026035198569297792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,16,power_law_1.2,0.10241279602050782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.01,0.34267520904541016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,16,power_law_1.01,0.03429119884967804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,16,power_law_1.2,0.1023103952407837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.01,0.4828159809112549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,16,power_law_1.01,0.04806399941444397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,16,power_law_1.2,0.10408320426940917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.01,0.6125440120697021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,16,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,16,power_law_1.2,0.10482560396194458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,16,power_law_1.01,0.7003903865814209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,16,power_law_1.01,0.06108160018920898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,16,power_law_1.2,0.10701440572738648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,16,power_law_1.01,0.06219519972801209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,16,power_law_1.2,0.10008959770202637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,16,power_law_1.01,1.2646464347839355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,16,power_law_1.01,0.06229119896888733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,16,power_law_1.2,0.10502400398254394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,16,power_law_1.01,2.339481544494629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,16,power_law_1.01,0.07114880084991455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,16,power_law_1.2,0.10529919862747192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,16,power_law_1.01,0.06753280162811279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,16,power_law_1.2,0.10859520435333252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,16,power_law_1.01,0.06690559983253479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,16,power_law_1.2,0.11452159881591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,16,power_law_1.01,0.06631039977073669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,16,power_law_1.2,0.11715199947357177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,16,power_law_1.01,0.0646336019039154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,16,power_law_1.2,0.1296512007713318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,16,power_law_1.01,0.06715520024299622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,16,power_law_1.2,0.14370559453964232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,16,power_law_1.01,0.06286079883575439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,16,power_law_1.2,0.17621760368347167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.01,0.07164160013198853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,16,power_law_1.2,0.21933441162109374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.01,0.08047360181808472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,16,power_law_1.2,0.2832191944122314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.01,0.0912447988986969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,16,power_law_1.2,0.3170880079269409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,16,power_law_1.2,0.4242432117462158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.01,0.11555839776992798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.01,0.13374719619750977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,16,power_law_1.2,0.6041344165802002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.01,0.19426560401916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,16,power_law_1.2,0.7789311885833741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.01,0.2482815980911255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,16,power_law_1.2,1.0996416091918946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.01,0.3442368030548096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,16,power_law_1.2,1.448960018157959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.01,0.469708776473999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,16,power_law_1.2,2.2745023727416993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,16,power_law_1.01,0.5266496181488037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,16,power_law_1.2,5.35667839050293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,16,power_law_1.01,0.896025562286377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,16,power_law_1.01,1.6755456924438477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,16,power_law_1.2,0.06824960112571717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,16,power_law_1.2,0.042777600884437564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,16,power_law_1.2,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,16,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,16,power_law_1.2,0.057011198997497556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,16,power_law_1.2,0.0673792004585266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,16,power_law_1.2,0.06710399985313416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,16,power_law_1.2,0.07344639897346497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,16,power_law_1.2,0.07184640169143677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,16,power_law_1.2,0.07200639843940734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,16,power_law_1.2,0.07455999851226806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,16,power_law_1.2,0.07455360293388366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,16,power_law_1.2,0.07604479789733887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,16,power_law_1.2,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,16,power_law_1.2,0.08119040131568908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,16,power_law_1.2,0.08431360125541687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,16,power_law_1.2,0.07368320226669312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,16,power_law_1.2,0.08740479946136474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,16,power_law_1.2,0.10306559801101685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,16,power_law_1.2,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,16,power_law_1.2,0.11223039627075196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,16,power_law_1.2,0.048467200994491574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,16,power_law_1.2,0.13507200479507447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,16,power_law_1.2,0.05779839754104614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,16,power_law_1.2,0.15545599460601806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,16,power_law_1.2,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,16,power_law_1.2,0.07081599831581116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,16,power_law_1.2,0.2111680030822754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,16,power_law_1.2,0.06943359971046448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,16,power_law_1.2,0.29173119068145753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,16,power_law_1.2,0.0711359977722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,16,power_law_1.2,0.3495039939880371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,16,power_law_1.2,0.07031679749488831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,16,power_law_1.2,0.5001215934753418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,16,power_law_1.2,0.07092480063438415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,16,power_law_1.2,0.7638847827911377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,16,power_law_1.2,0.07073919773101807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,16,power_law_1.2,0.8824383735656738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,16,power_law_1.2,0.07214080095291138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,16,power_law_1.2,1.1337599754333496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,16,power_law_1.2,0.07192320227622986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,16,power_law_1.2,1.9398719787597656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,16,power_law_1.2,0.07294719815254211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,16,power_law_1.2,0.07837439775466919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,16,power_law_1.2,3.8383296966552733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,16,power_law_1.2,0.0815168023109436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,16,power_law_1.2,0.02645759880542755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,8,balanced,0.040474665661652885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,16,power_law_1.2,0.09175040125846863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,16,power_law_1.2,0.022291199862957002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,8,balanced,0.049413333336512245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,16,power_law_1.2,0.0972927987575531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,8,balanced,0.0532533327738444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,16,power_law_1.2,0.022163200378417968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,8,balanced,0.07146133482456207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.2,0.10921599864959716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,16,power_law_1.2,0.023379200696945192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,8,balanced,0.10479467113812764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,16,power_law_1.2,0.02563199996948242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.2,0.1372032046318054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,8,balanced,0.09918933113416036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,16,power_law_1.2,0.03340159952640533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.2,0.15389440059661866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,8,balanced,0.11274666587511699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,16,power_law_1.2,0.04177919924259186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.2,0.19913599491119385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,8,balanced,0.10055466492970784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,16,power_law_1.2,0.04423039853572845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,8,balanced,0.10455999771753947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.2,0.23311998844146728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,8,balanced,0.10332266489664714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,16,power_law_1.2,0.05910400152206421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.2,0.3799040079116821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,8,balanced,0.09496000409126282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,16,power_law_1.2,0.05940480232238769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.2,0.404307222366333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,8,balanced,0.10062932968139648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,16,power_law_1.2,0.059680002927780154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.2,0.6695168018341064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,8,balanced,0.0981760025024414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,16,power_law_1.2,0.0720192015171051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.2,0.7649216175079345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,16,power_law_1.2,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,8,balanced,0.09588799873987834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,16,power_law_1.2,1.2523839950561524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,16,power_law_1.2,0.06661760210990905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,8,balanced,0.10297600428263347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,16,power_law_1.2,0.05973759889602661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,16,power_law_1.2,1.5216383934020996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,8,balanced,0.10379200180371602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,16,power_law_1.2,0.0617792010307312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,8,balanced,0.10359467069307964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,16,power_law_1.2,3.689228820800781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,16,power_law_1.2,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,8,balanced,0.10779733459154765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,16,power_law_1.2,0.06043519973754883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.2,0.06837760210037232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,8,balanced,0.10963732997576396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.2,0.0853056013584137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.2,0.09427840113639832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,8,balanced,0.1183733344078064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.2,0.11559040546417236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.2,0.15393919944763185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,8,balanced,0.12571733196576437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.2,0.208076810836792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.2,0.28932480812072753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.2,0.3778496026992798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,8,balanced,0.15636266271273294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.2,0.5638527870178223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,16,power_law_1.2,0.6373631954193115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,16,power_law_1.2,1.1771648406982422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,8,balanced,0.16496533155441284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,16,power_law_1.2,2.2352832794189452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,8,balanced,0.21383466323216757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,8,balanced,0.2651573419570923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,8,balanced,0.3585546811421712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,8,balanced,0.46750934918721515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,8,balanced,0.5661919911702474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,8,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,8,balanced,0.04053866614898046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,8,balanced,0.041237334410349526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,8,balanced,0.053173333406448364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,8,balanced,0.07134399811426799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,8,balanced,0.07303466896216075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,8,balanced,0.07527466615041097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,8,balanced,0.07365866502126057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,8,balanced,0.07161599894364674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,8,balanced,0.07534400125344594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,8,balanced,0.07153599957625072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,8,balanced,0.07605866591135661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,8,balanced,0.0740533322095871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,8,balanced,0.07375466823577881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,8,balanced,0.08250666658083598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,8,balanced,0.07985599835713704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,8,balanced,0.8711360295613607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,8,balanced,0.08346133430798848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,8,balanced,0.08986666798591614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,8,balanced,0.09870933492978413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,8,balanced,0.10705600182215373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,8,balanced,0.12001066406567891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,8,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,8,balanced,0.043007999658584595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,8,balanced,0.15215999881426492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,8,balanced,0.05152533451716105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,8,balanced,0.056362668673197426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,8,balanced,0.07931200166543324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,8,balanced,0.07161066432793935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,8,balanced,0.17579734325408936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,8,balanced,0.06830933193365733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,8,balanced,0.0744053324063619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,8,balanced,0.07309866448243459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,8,balanced,0.07434133191903432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,8,balanced,0.07339733342329662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,8,balanced,0.23644800980885824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,8,balanced,0.07128533224264781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,8,balanced,0.079434668024381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,8,balanced,0.07018133501211803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,8,balanced,0.08250666658083598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,8,balanced,0.02139200021823247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,8,balanced,0.07911466558774312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,8,balanced,0.024175999065240223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,8,balanced,0.29604266087214154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,8,balanced,0.023973333338896435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,8,balanced,0.07803733150164287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,8,balanced,0.024656000236670177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,8,balanced,0.03159466634194056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,8,balanced,0.08299200236797333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,8,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,8,balanced,0.04158399999141693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,8,balanced,0.08820266524950664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,8,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,8,balanced,0.057904000083605446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,8,balanced,0.09711999694506328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,8,balanced,0.055829331278800964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,8,balanced,0.05236266553401947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,8,balanced,0.41921599706013996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,8,balanced,0.06713599960009257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,8,balanced,0.10519466797510783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,8,balanced,0.06880533198515575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,8,balanced,0.06503466765085857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,8,balanced,0.08180266618728638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,8,balanced,0.14256532986958823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,8,balanced,0.07812800010045369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,8,balanced,0.08056533336639404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,8,balanced,0.07996266583601634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,8,balanced,1.6808959643046062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,8,balanced,0.15657066305478415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,8,balanced,0.09372799595197041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,8,balanced,0.5398933490117391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,8,balanced,0.09875200192133586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,8,balanced,0.11245333154996236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,8,balanced,0.20660799741744995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,8,balanced,0.13063466548919678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,8,balanced,0.2553173303604126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,8,balanced,0.1490666667620341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,8,balanced,0.6683359940846761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,8,balanced,0.2233333388964335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,8,balanced,0.35688531398773193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,8,balanced,0.255568007628123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,8,balanced,0.4540799856185913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,8,balanced,0.3678933382034302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,8,balanced,1.0279573599497478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,8,balanced,0.4785173336664836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,8,balanced,0.5500640074412028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,8,balanced,0.5948053201039633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,8,balanced,0.8456693490346273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,8,balanced,1.9852693875630696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,8,balanced,0.9488053321838379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,8,power_law_1.01,0.06651520133018493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,8,power_law_1.01,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,8,power_law_1.01,0.05602560043334961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,8,power_law_1.01,0.06874880194664001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,8,power_law_1.01,0.07214720249176025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,8,power_law_1.01,0.0933184027671814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,8,power_law_1.01,0.09741439819335937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,8,power_law_1.01,0.09711999893188476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,8,power_law_1.01,0.09571840167045594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,8,power_law_1.01,0.10059520006179809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,8,power_law_1.01,0.1023103952407837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,8,power_law_1.01,0.10034559965133667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,8,power_law_1.01,0.09898239970207215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,8,power_law_1.01,0.1022271990776062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,8,balanced,1.6104480425516765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,8,power_law_1.01,0.10577280521392822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,8,power_law_1.01,0.10725120306015015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,8,power_law_1.01,0.11217279434204101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,8,balanced,1.8659626642862956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,8,power_law_1.01,0.11905280351638795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,8,power_law_1.01,0.1303936004638672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,8,power_law_1.01,0.15418239831924438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,8,power_law_1.01,0.1679744005203247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,8,power_law_1.01,0.21092479228973388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,8,power_law_1.01,0.26340479850769044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,8,power_law_1.01,0.32997119426727295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,8,power_law_1.01,0.4694975852966309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,8,power_law_1.01,0.5794623851776123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,8,power_law_1.01,0.7812928199768067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,8,power_law_1.01,0.9306431770324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,8,power_law_1.01,1.3989184379577637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,8,power_law_1.01,2.719526481628418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,8,power_law_1.01,0.053235197067260744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,8,power_law_1.01,0.04217599928379059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,8,power_law_1.01,0.045817598700523376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,8,power_law_1.01,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,8,power_law_1.01,0.05632640123367309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,8,power_law_1.01,0.07006719708442688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,8,power_law_1.01,0.07212799787521362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,8,power_law_1.01,0.07288960218429566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,8,power_law_1.01,0.07233279943466187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,8,power_law_1.01,0.0741823971271515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,8,power_law_1.01,0.07417600154876709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,8,power_law_1.01,0.07439360022544861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,8,power_law_1.01,0.07581440210342408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,8,power_law_1.01,0.0769599974155426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,8,power_law_1.01,0.08188160061836243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,8,power_law_1.01,0.08482559919357299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,8,power_law_1.01,0.08643199801445008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,8,power_law_1.01,0.09658240079879761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,8,power_law_1.01,0.10551680326461792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,8,power_law_1.01,0.12821120023727417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,8,power_law_1.01,0.1404736042022705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,8,power_law_1.01,0.19088640213012695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,8,power_law_1.01,0.23138558864593506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,8,power_law_1.01,0.330182409286499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,8,power_law_1.01,0.4291520118713379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,8,power_law_1.01,0.5956992149353028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,8,power_law_1.01,0.7654528141021728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,8,power_law_1.01,0.9426112174987793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,8,power_law_1.01,1.4967167854309082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,8,power_law_1.01,3.0261760711669923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,8,power_law_1.01,0.021491199731826782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,8,power_law_1.01,0.021376000344753267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,8,power_law_1.01,0.061561602354049685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,8,power_law_1.01,0.022303999960422517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,8,power_law_1.01,0.0538752019405365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,8,power_law_1.01,0.022329600155353548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,8,power_law_1.01,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,8,power_law_1.01,0.024716800451278685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,8,power_law_1.01,0.0546176016330719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,8,power_law_1.01,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,8,power_law_1.01,0.05720959901809693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,8,power_law_1.01,0.036083200573921205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,8,power_law_1.01,0.07218559980392455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,8,power_law_1.01,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,8,power_law_1.01,0.0707647979259491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,8,power_law_1.01,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,8,power_law_1.01,0.07265920042991639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,8,power_law_1.01,0.044870400428771974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,8,power_law_1.01,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,8,power_law_1.01,0.046726399660110475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,8,power_law_1.01,0.07303680181503296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,8,power_law_1.01,0.06295040249824524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,8,power_law_1.01,0.07550719976425171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,8,power_law_1.01,0.0632960021495819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,8,power_law_1.01,0.07642239928245545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,8,power_law_1.01,0.06484479904174804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,8,power_law_1.01,0.07512959837913513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,8,power_law_1.01,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,8,power_law_1.01,0.07713279724121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,8,power_law_1.01,0.06874240040779114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,8,power_law_1.01,0.08053119778633118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,8,power_law_1.01,0.06583679914474487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,8,power_law_1.01,0.0861952006816864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,8,power_law_1.01,0.0664896011352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,8,power_law_1.01,0.08857600092887878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.01,0.06291840076446534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,8,power_law_1.01,0.09664639830589294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.01,0.0766592025756836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.01,0.10880000591278076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.01,0.08507519960403442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,8,power_law_1.2,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.01,0.10368640422821045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,8,power_law_1.2,0.051923197507858274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.01,0.1193343997001648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,8,power_law_1.2,0.05482879877090454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.01,0.16136959791183472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,8,power_law_1.2,0.06179839968681335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.01,0.1916416049003601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,8,power_law_1.2,0.07189760208129883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.01,0.2720896005630493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,8,power_law_1.2,0.0944703996181488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.01,0.36664319038391113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,8,power_law_1.2,0.10188159942626954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,8,power_law_1.01,0.43232002258300783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,8,power_law_1.2,0.09857280254364013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,8,power_law_1.01,0.6920512199401856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,8,power_law_1.2,0.10096640586853027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,8,power_law_1.2,0.10218240022659301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,8,power_law_1.01,1.4634431838989257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,8,power_law_1.2,0.10291839838027954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,8,power_law_1.2,0.10146559476852417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,8,power_law_1.2,0.10058239698410035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,8,power_law_1.2,0.10213119983673095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,8,power_law_1.2,0.1080191969871521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,8,power_law_1.2,0.11344000101089477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,8,power_law_1.2,0.11207040548324584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,8,power_law_1.2,0.12492799758911133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,8,power_law_1.2,0.13586560487747193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,8,power_law_1.2,0.15459840297698973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,8,power_law_1.2,0.17315200567245484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,8,power_law_1.2,0.22573440074920653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,8,power_law_1.2,0.2873663902282715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,8,power_law_1.2,0.3533184051513672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,8,power_law_1.2,0.5110527992248535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,8,power_law_1.2,0.6951039791107178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,8,power_law_1.2,0.8934144020080567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,8,power_law_1.2,1.1025983810424804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,8,power_law_1.2,1.8612031936645508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,8,power_law_1.2,3.653696060180664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,8,power_law_1.2,0.04844799935817719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,8,power_law_1.2,0.042124798893928526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,8,power_law_1.2,0.04620159864425659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,8,power_law_1.2,0.049958398938179015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,8,power_law_1.2,0.05255039930343628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,8,power_law_1.2,0.06871680021286011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,8,power_law_1.2,0.07312639951705932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,8,power_law_1.2,0.07336320281028748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,8,power_law_1.2,0.07154560089111328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,8,power_law_1.2,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,8,power_law_1.2,0.07590399980545044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,8,power_law_1.2,0.07354239821434021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,8,power_law_1.2,0.07578240036964416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,8,power_law_1.2,0.07747200131416321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,8,power_law_1.2,0.08369920253753663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,8,power_law_1.2,0.08594560027122497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,8,power_law_1.2,0.08965759873390197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,8,power_law_1.2,0.10033919811248779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,8,power_law_1.2,0.11192959547042847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,8,power_law_1.2,0.1323456048965454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,8,power_law_1.2,0.15835520029067993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,8,power_law_1.2,0.19319679737091064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,8,power_law_1.2,0.26415998935699464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,8,power_law_1.2,0.3553407907485962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,8,power_law_1.2,0.5401535987854004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,8,power_law_1.2,0.6662144184112548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,8,power_law_1.2,0.9829952239990234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,8,power_law_1.2,1.2089983940124511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,8,power_law_1.2,0.06041600108146668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,8,power_law_1.2,1.8380992889404297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,8,power_law_1.2,0.043609601259231565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,8,power_law_1.2,0.05086719989776611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,8,power_law_1.2,4.895289611816406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,8,power_law_1.2,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,8,power_law_1.2,0.05565440058708191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,8,power_law_1.2,0.07207679748535156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,8,power_law_1.2,0.07438079714775085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,8,power_law_1.2,0.07245439887046815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,8,power_law_1.2,0.07408000230789184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,8,power_law_1.2,0.07553920149803162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.01,0.12188160419464111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,8,power_law_1.2,0.07391999959945679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.01,0.1463104009628296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,8,power_law_1.2,0.07355520129203796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.01,0.17326079607009887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,8,power_law_1.2,0.07772160172462464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.01,0.22480640411376954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,8,power_law_1.2,0.07616639733314515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.01,0.28589439392089844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,8,power_law_1.2,0.08264319896697998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.01,0.3796159982681274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,8,power_law_1.2,0.0874239981174469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.01,0.509497594833374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,8,power_law_1.2,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.01,0.6500160217285156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,8,power_law_1.2,0.09777920246124268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,8,power_law_1.01,0.8679936408996582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.2,0.1095039963722229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,8,power_law_1.01,1.2479488372802734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.2,0.13518719673156737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.2,0.15133440494537354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,8,power_law_1.01,2.3702976226806642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,8,power_law_1.2,0.021721599996089934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.2,0.19842560291290284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,8,power_law_1.2,0.021484799683094025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.2,0.23991680145263672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,8,power_law_1.2,0.020032000541687012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.2,0.308025598526001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,8,power_law_1.2,0.02263039946556091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,8,power_law_1.2,0.026009601354599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.2,0.3964927911758423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,8,power_law_1.2,0.029497599601745604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.2,0.7091584205627441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,8,power_law_1.2,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.2,0.8627391815185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,8,power_law_1.2,0.03427839875221252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,8,power_law_1.2,1.1771967887878418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,8,power_law_1.2,0.0471807986497879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,8,power_law_1.2,1.708652877807617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,8,power_law_1.2,0.04650239944458008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,8,power_law_1.2,0.044921600818634035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,8,power_law_1.2,3.620217514038086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,8,power_law_1.2,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,8,power_law_1.2,0.06348800063133239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,8,power_law_1.2,0.06562560200691223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,8,power_law_1.2,0.0692351996898651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,8,power_law_1.2,0.06521599888801574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,8,power_law_1.2,0.06867200136184692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,8,power_law_1.2,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.2,0.06878079771995545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.2,0.07708799839019775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.2,0.0884607970714569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.2,0.11045119762420655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.2,0.12865920066833497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.2,0.17843199968338014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.2,0.21059200763702393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.2,0.32196478843688964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.2,0.4031871795654297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,8,power_law_1.2,0.4763328075408936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,8,power_law_1.2,0.7705344200134278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,8,power_law_1.2,1.5362303733825684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,4,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,4,balanced,0.04520533482233683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,4,balanced,0.051728000243504844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,4,balanced,0.0717439999183019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,4,balanced,0.10272533694903056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,4,balanced,0.09846400221188863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,4,balanced,0.09846933682759602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,4,balanced,0.09850133458773296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,4,balanced,0.09795733292897542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,4,balanced,0.09787733356157939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,4,balanced,0.09852799773216248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,4,balanced,0.09826667110125224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,4,balanced,0.10097600022951762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,4,balanced,0.10020266969998677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,4,balanced,0.10408000151316325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,4,balanced,0.10586667060852051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,4,balanced,0.11248532931009929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,4,balanced,0.11507733662923177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,4,balanced,0.11942399541536967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,4,balanced,0.13219199577967325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,4,balanced,0.1416106621424357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,4,balanced,0.03828799972931544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,4,balanced,0.04077333211898804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,4,balanced,0.043162668744723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,4,balanced,0.042463997999827065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,4,balanced,0.17968533436457315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,4,balanced,0.05342400074005127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,4,balanced,0.04550399879614512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,4,balanced,0.08027199904123943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,4,balanced,0.053455998500188194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,4,balanced,0.07999999821186066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,4,balanced,0.06354133288065593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,4,balanced,0.07895466685295105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,4,balanced,0.09057066837946574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,4,balanced,0.2051573395729065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,4,balanced,0.07974933087825775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,4,balanced,0.08672533432642619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,4,balanced,0.0778186668952306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,4,balanced,0.08590400218963623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,4,balanced,0.0774239997069041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,4,balanced,0.08653333783149719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,4,balanced,0.07844266792138417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,4,balanced,0.08494399984677632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,4,balanced,0.08169599870840709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,4,balanced,0.08692266543706258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,4,balanced,0.0817333310842514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,4,balanced,0.07928533355395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,4,balanced,0.2728319962819417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,4,balanced,0.08447466293970744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,4,balanced,0.08220799763997395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,4,balanced,0.08793600400288899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,4,balanced,0.08999466896057129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,4,balanced,0.08359466989835103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,4,balanced,0.08808533350626628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,4,balanced,0.0936853289604187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,4,balanced,0.09197866916656494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,4,balanced,0.09231467048327129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,4,balanced,0.10260267059008281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,4,balanced,0.0960106650988261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,4,balanced,0.3448266585667928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,4,balanced,0.10373866558074951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,4,balanced,0.10843200484911601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,4,balanced,0.12635733683904013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,4,balanced,0.14089066783587137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,4,balanced,0.47627735137939453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,4,balanced,0.021642667551835377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,4,balanced,0.19633066654205322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,4,balanced,0.022848000129063923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,4,balanced,0.023792001108328503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,4,balanced,0.024447999894618988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,4,balanced,0.03139200061559677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,4,balanced,0.222053329149882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,4,balanced,0.03291733314593633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,4,balanced,0.035173334181308746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,4,balanced,0.03655466685692469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,4,balanced,0.043237333496411644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,4,balanced,0.042853335539499916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,4,balanced,0.042650664846102394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,4,balanced,0.3085493246714274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,4,balanced,0.05749866863091787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,4,balanced,0.058037335673967995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,4,balanced,0.05672533313433329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,4,balanced,0.6154346863428751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,4,balanced,0.07331733405590057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,4,balanced,0.38521599769592285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,4,balanced,0.07328000168005626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,4,balanced,0.07386666536331177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,4,balanced,0.08960533142089844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,4,balanced,0.09718933701515198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,4,balanced,0.10753066341082256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,4,balanced,0.5470453500747681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,4,balanced,0.11744532982508342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,4,balanced,0.1295253336429596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,4,balanced,0.7441066900889078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,4,balanced,0.13608533143997192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,4,balanced,0.7101013660430908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,4,balanced,0.21450134118398032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,4,balanced,0.23668799797693887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,4,balanced,0.8738666375478109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,4,balanced,1.1516640186309814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,4,balanced,0.3383306662241618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,4,balanced,0.4411626656850179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,4,balanced,1.3485760688781738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,4,balanced,0.5386240084966024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,4,balanced,2.224506696065267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,4,balanced,0.8637066682179769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,4,balanced,0.11026133100191753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,4,balanced,2.6012372970581055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,4,balanced,0.12431466579437256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,4,balanced,0.1407360037167867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,4,balanced,0.18972800175348917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,4,balanced,0.22113066911697388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,4,balanced,0.30515732367833454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,4,balanced,0.39082666238149005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,4,balanced,0.5581173499425253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,4,balanced,1.687930742899577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,4,balanced,0.7225279808044434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,4,balanced,0.8907519976298014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,4,power_law_1.01,0.05059199929237366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,4,power_law_1.01,0.05079039931297302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,4,power_law_1.01,0.051481598615646364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,4,power_law_1.01,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,4,power_law_1.01,0.06914560198783874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,4,balanced,1.3800320625305176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,4,power_law_1.01,0.09523839950561523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,4,power_law_1.01,0.09767680168151856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,4,power_law_1.01,0.09808639883995056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,4,power_law_1.01,0.09559040069580078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,4,power_law_1.01,0.09751039743423462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,4,power_law_1.01,0.09963520169258118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,4,power_law_1.01,0.10085760354995728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,4,power_law_1.01,0.10066560506820679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,4,power_law_1.01,0.10170880556106568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,4,power_law_1.01,0.04589439928531647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,4,power_law_1.01,0.10763520002365112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,4,power_law_1.01,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,4,power_law_1.01,0.10856959819793702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,4,power_law_1.01,0.044761601090431216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,4,power_law_1.01,0.11296639442443848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,4,power_law_1.01,0.054150402545928955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,4,power_law_1.01,0.11963520050048829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,4,power_law_1.01,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,4,power_law_1.01,0.1325055956840515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,4,power_law_1.01,0.07031679749488831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,4,power_law_1.01,0.07459840178489685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,4,power_law_1.01,0.15656960010528564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,4,power_law_1.01,0.0763647973537445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,4,power_law_1.01,0.17067519426345826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,4,power_law_1.01,0.07630720138549804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,4,power_law_1.01,0.2224639892578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,4,power_law_1.01,0.07550719976425171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,4,power_law_1.01,0.25754880905151367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,4,power_law_1.01,0.07899519801139832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,4,power_law_1.01,0.34482560157775877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,4,power_law_1.01,0.08058239817619324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,4,power_law_1.01,0.43137922286987307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,4,power_law_1.01,0.08205440044403076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,4,power_law_1.01,0.5941823959350586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,4,power_law_1.01,0.08386560082435608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,4,power_law_1.01,0.8099776268005371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,4,power_law_1.01,0.09084799885749817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,4,power_law_1.01,0.9186559677124023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,4,power_law_1.01,0.09237759709358215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,4,power_law_1.01,1.3737471580505372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,4,power_law_1.01,0.09568639993667602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,4,power_law_1.01,0.10670720338821411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,4,power_law_1.01,2.9754432678222655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,4,power_law_1.01,0.11941759586334229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,4,power_law_1.01,0.1478976011276245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,4,power_law_1.01,0.1725119948387146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,4,power_law_1.01,0.21827199459075927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,4,power_law_1.01,0.2726016044616699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,4,power_law_1.01,0.35980799198150637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,4,power_law_1.01,0.49280638694763185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,4,power_law_1.01,0.6906816005706787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,4,power_law_1.01,0.9126079559326172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,4,power_law_1.01,1.046675205230713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,4,power_law_1.01,1.866476821899414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,4,balanced,2.6743361155192056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,4,power_law_1.01,0.055212801694869994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,4,power_law_1.01,3.515411376953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,4,power_law_1.01,0.04498560130596161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,4,power_law_1.01,0.04604800045490265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,4,power_law_1.01,0.061388802528381345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,4,power_law_1.01,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,4,power_law_1.01,0.08367999792098998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,4,power_law_1.01,0.08462079763412475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,4,power_law_1.01,0.08542720079421998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,4,power_law_1.01,0.08592000007629394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,4,power_law_1.01,0.08848639726638793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,4,power_law_1.01,0.08640639781951905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,4,power_law_1.01,0.08604159951210022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,4,power_law_1.01,0.08822399973869324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,4,power_law_1.01,0.09029120206832886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,4,power_law_1.01,0.09484800100326538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,4,power_law_1.01,0.09662079811096191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,4,power_law_1.01,0.10409599542617798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,4,power_law_1.01,0.11418240070343018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.01,0.13509759902954102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.01,0.15192960500717162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.01,0.18213119506835937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.01,0.23327360153198243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.01,0.2636735916137695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.01,0.4220287799835205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.01,0.4576064109802246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.01,0.6991744041442871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.01,0.8635392189025879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,4,power_law_1.01,1.1190272331237794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,4,power_law_1.01,1.686240005493164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,4,power_law_1.01,3.291097640991211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,4,power_law_1.01,0.01863040030002594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,4,power_law_1.01,0.01971839964389801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,4,power_law_1.01,0.021049599349498748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,4,power_law_1.01,0.022777600586414336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,4,power_law_1.01,0.024659200012683867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,4,power_law_1.01,0.028838399052619933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,4,power_law_1.01,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,4,power_law_1.01,0.030854400992393494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,4,power_law_1.01,0.03501439988613129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,4,power_law_1.01,0.03529599905014038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,4,power_law_1.2,0.05186560153961182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,4,power_law_1.01,0.03437440097332001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,4,power_law_1.2,0.04865280091762543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,4,power_law_1.01,0.043270400166511534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,4,power_law_1.2,0.04875519871711731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,4,power_law_1.01,0.04431360065937042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,4,power_law_1.2,0.04601599872112274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,4,power_law_1.2,0.05999360084533691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,4,power_law_1.2,0.041145598888397215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,4,power_law_1.01,0.0427264004945755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,4,power_law_1.2,0.06749439835548401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,4,power_law_1.2,0.04500479996204376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,4,power_law_1.01,0.06796159744262695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,4,power_law_1.2,0.0934656023979187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,4,power_law_1.2,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,4,power_law_1.2,0.0997376024723053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,4,power_law_1.01,0.06788480281829834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,4,power_law_1.2,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,4,power_law_1.2,0.101254403591156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,4,power_law_1.01,0.07095680236816407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,4,power_law_1.2,0.07626240253448487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,4,power_law_1.2,0.10039039850234985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,4,power_law_1.01,0.07126399874687195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,4,power_law_1.2,0.07596799731254578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,4,power_law_1.2,0.09926400184631348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.01,0.0729088008403778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,4,power_law_1.2,0.07730559706687927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,4,power_law_1.2,0.1011072039604187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.01,0.08499839901924133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,4,power_law_1.2,0.07609599828720093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,4,power_law_1.2,0.10079360008239746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,4,power_law_1.2,0.0779263973236084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,4,power_law_1.2,0.09455999732017517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.01,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,4,power_law_1.2,0.10488320589065551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,4,power_law_1.2,0.07740799784660339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.01,0.10927360057830811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,4,power_law_1.2,0.11130880117416382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,4,power_law_1.2,0.08195840120315552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.01,0.1296895980834961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,4,power_law_1.2,0.11013120412826538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,4,power_law_1.2,0.08113920092582702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.01,0.16549760103225708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,4,power_law_1.2,0.11504000425338745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,4,power_law_1.2,0.08414080142974853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.01,0.20762879848480226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,4,power_law_1.2,0.12647680044174195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,4,power_law_1.2,0.09094399809837342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,4,power_law_1.2,0.1358847975730896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,4,power_law_1.2,0.0934719979763031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.01,0.28132479190826415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,4,power_law_1.2,0.16021759510040284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,4,power_law_1.2,0.09788159728050232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.01,0.3588992118835449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,4,power_law_1.2,0.17836159467697144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,4,power_law_1.2,0.11178239583969116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,4,power_law_1.01,0.4367231845855713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,4,power_law_1.2,0.23147521018981934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,4,power_law_1.2,0.12449920177459717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,4,power_law_1.2,0.2912447929382324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,4,power_law_1.2,0.1537727952003479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,4,power_law_1.01,0.6854080200195313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,4,power_law_1.2,0.1764415979385376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,4,power_law_1.2,0.35927040576934816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,4,power_law_1.01,1.2912063598632812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,4,power_law_1.2,0.2228287935256958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,4,power_law_1.2,0.47679362297058103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,4,power_law_1.2,0.285318398475647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,4,power_law_1.2,0.7628799915313721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,4,power_law_1.2,0.4024960041046143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,4,power_law_1.2,0.9924863815307617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,4,power_law_1.2,0.529420804977417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,4,power_law_1.2,1.086796760559082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,4,power_law_1.2,0.8071743965148925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,4,power_law_1.2,1.727052879333496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,4,power_law_1.2,1.0104191780090332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,4,power_law_1.2,3.793356704711914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,4,power_law_1.2,1.1775872230529785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,4,power_law_1.2,2.0357183456420898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,4,power_law_1.2,4.383168029785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,4,power_law_1.2,0.05111039876937866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,4,power_law_1.2,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,4,power_law_1.2,0.05208960175514221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,4,power_law_1.2,0.05767679810523987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,4,power_law_1.2,0.06591359972953796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,4,power_law_1.2,0.08250880241394043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,4,power_law_1.2,0.08464639782905578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,4,power_law_1.2,0.08729599714279175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,4,power_law_1.2,0.08978559970855712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,4,power_law_1.2,0.08789759874343872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,4,power_law_1.2,0.08615679740905761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,4,power_law_1.2,0.08817920088768005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,4,power_law_1.2,0.08731520175933838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,4,power_law_1.2,0.09251199960708618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,4,power_law_1.2,0.09949439764022827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,4,power_law_1.2,0.10408960580825806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,4,power_law_1.2,0.10870399475097656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,4,power_law_1.2,0.12233599424362182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.2,0.13266559839248657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.2,0.15991679430007935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.2,0.19145599603652955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.2,0.2277440071105957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.2,0.27654399871826174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.2,0.3819391965866089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.2,0.5061888217926025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.2,0.879635238647461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.2,1.1667327880859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,4,power_law_1.2,1.3145024299621582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,4,power_law_1.2,2.1257471084594726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,4,power_law_1.2,4.947782516479492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,4,power_law_1.2,0.019232000410556793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,4,power_law_1.2,0.019244800508022308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,4,power_law_1.2,0.021408000588417055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,4,power_law_1.2,0.022361600399017335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,4,power_law_1.2,0.023622399568557738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,2,balanced,0.040106666584809623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,2,balanced,0.04409599800904592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,4,power_law_1.2,0.029311999678611755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,2,balanced,0.046485334634780884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,2,balanced,0.046949331959088646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,2,balanced,0.04867733518282572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,4,power_law_1.2,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,2,balanced,0.055685331424077354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,2,balanced,0.06399466594060262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,4,power_law_1.2,0.031763198971748355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,2,balanced,0.07482133309046428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,2,balanced,0.09498133261998494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,4,power_law_1.2,0.03555200099945068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,2,balanced,0.10915733377138774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,2,balanced,0.09285333752632141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,2,balanced,0.11312533418337505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,4,power_law_1.2,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,2,balanced,0.09227733810742696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,2,balanced,0.10573866963386536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,4,power_law_1.2,0.035206401348114015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,2,balanced,0.09184533357620239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,2,balanced,0.1092639962832133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,4,power_law_1.2,0.04389120042324066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,2,balanced,0.09583466251691182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,2,balanced,0.1090186635653178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,2,balanced,0.0936853289604187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,4,power_law_1.2,0.04260480105876922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,2,balanced,0.10839999715487163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,2,balanced,0.09403199950853984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,4,power_law_1.2,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,2,balanced,0.10703999797503154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,2,balanced,0.09618666768074036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,4,power_law_1.2,0.0686464011669159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,2,balanced,0.11231999595959981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,2,balanced,0.09830400347709656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,2,balanced,0.11008000373840332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,4,power_law_1.2,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,2,balanced,0.11306132872899373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,4,power_law_1.2,0.07107840180397033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,2,balanced,0.09959999720255534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,4,power_law_1.2,0.0741375982761383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,2,balanced,0.11921067039171855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,2,balanced,0.1042133371035258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.2,0.07909759879112244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,2,balanced,0.11859200398127238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,2,balanced,0.11033067107200623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.2,0.0841152012348175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,2,balanced,0.12326400478680928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,2,balanced,0.11390933394432068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.2,0.09484800100326538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,2,balanced,0.13557333747545877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,2,balanced,0.12897599736849466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.2,0.11048959493637085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.2,0.12865920066833497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,2,balanced,0.14588266611099243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,2,balanced,0.13913599650065103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.2,0.17244160175323486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.2,0.2085952043533325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,2,balanced,0.16172266999880472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,2,balanced,0.1661066710948944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.2,0.2802112102508545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.2,0.3574847936630249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,2,balanced,0.17890665928522745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,2,balanced,0.1857759952545166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,4,power_law_1.2,0.46796159744262694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,4,power_law_1.2,0.690124797821045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,2,balanced,0.25045865774154663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,2,balanced,0.26815466086069745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,4,power_law_1.2,1.3897600173950195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,2,balanced,0.2930293281873067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,2,balanced,0.3167733351389567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,2,balanced,0.4031413396199544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,2,balanced,0.4540853500366211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,2,balanced,0.5916426579157511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,2,balanced,0.5151626666386923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,2,power_law_1.01,0.04683519899845123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,2,power_law_1.01,0.04836480021476745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,2,power_law_1.01,0.05304960012435913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,2,power_law_1.01,0.06698240041732788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,2,power_law_1.01,0.07452800273895263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,2,power_law_1.01,0.10295039415359497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,2,balanced,0.8530560334523519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,2,balanced,0.7162453333536783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,2,power_law_1.01,0.10188800096511841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,2,power_law_1.01,0.10384639501571655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,2,power_law_1.01,0.10335999727249146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,2,power_law_1.01,0.10581760406494141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,2,power_law_1.01,0.10811519622802734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,2,power_law_1.01,0.1123263955116272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,2,power_law_1.01,0.11498880386352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,2,power_law_1.01,0.11311999559402466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,2,power_law_1.01,0.12008960247039795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,2,power_law_1.01,0.12400000095367432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,2,balanced,1.1095840136210124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,2,balanced,0.9213866392771403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,2,power_law_1.01,0.126803195476532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,2,power_law_1.01,0.14231679439544678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,2,power_law_1.01,0.15240319967269897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,2,power_law_1.01,0.18424960374832153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,2,power_law_1.01,0.2128063917160034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,2,power_law_1.01,0.2790080070495605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,2,power_law_1.01,0.3233151912689209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,2,power_law_1.01,0.43517441749572755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,2,power_law_1.01,0.5509759902954101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,2,power_law_1.01,0.7747519969940185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,2,power_law_1.01,1.006553554534912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,2,power_law_1.01,1.2786623954772949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,2,balanced,1.3652159372965496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,2,balanced,1.127061367034912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,2,power_law_1.01,2.091129684448242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,2,power_law_1.01,3.857235336303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,2,balanced,2.1438825925191245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,2,balanced,1.7489013671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,2,power_law_1.01,0.04318720102310181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,2,power_law_1.01,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,2,power_law_1.01,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,2,power_law_1.01,0.05725439786911011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,2,power_law_1.01,0.06257920265197754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,2,power_law_1.01,0.08354560136795045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,2,power_law_1.01,0.08592640161514283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,2,power_law_1.01,0.08794879913330078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,2,power_law_1.01,0.08809599876403809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,2,power_law_1.01,0.08970879912376403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,2,power_law_1.01,0.09212160110473633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,2,power_law_1.01,0.09567999839782715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,2,power_law_1.01,0.0950655996799469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,2,power_law_1.01,0.09800320267677307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,2,power_law_1.01,0.10643839836120605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,2,power_law_1.01,0.11021440029144287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,2,balanced,4.2004852294921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,2,power_law_1.01,0.11811840534210205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,2,balanced,3.4023412068684897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,2,power_law_1.01,0.13210879564285277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,2,power_law_1.01,0.1492735981941223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,2,power_law_1.01,0.17864320278167725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,2,power_law_1.01,0.21401600837707518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,2,power_law_1.01,0.28378241062164306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,2,power_law_1.01,0.37678720951080324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,2,power_law_1.01,0.4911231994628906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,2,power_law_1.01,0.6917183876037598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,2,power_law_1.01,0.9215680122375488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,2,power_law_1.01,1.1783807754516602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,2,power_law_1.01,1.487065601348877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,2,power_law_1.01,2.3597951889038087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,2,power_law_1.01,4.564704132080078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,2,power_law_1.2,0.046963199973106384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,2,power_law_1.2,0.04744960069656372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,2,power_law_1.2,0.053388798236846925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,2,power_law_1.2,0.0610368013381958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,2,power_law_1.2,0.06968960165977478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,2,power_law_1.2,0.09965440034866332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,2,power_law_1.2,0.10371199846267701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,2,power_law_1.2,0.10680960416793824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,2,power_law_1.2,0.10746239423751831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,2,power_law_1.2,0.10647039413452149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,2,power_law_1.2,0.11121280193328857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,2,power_law_1.2,0.11286400556564331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,2,power_law_1.2,0.11340800523757935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,2,power_law_1.2,0.11663360595703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,2,power_law_1.2,0.12280960083007812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,2,power_law_1.2,0.12414720058441162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,2,power_law_1.2,0.12999039888381958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,2,power_law_1.2,0.14493440389633178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,2,power_law_1.2,0.15890560150146485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,2,power_law_1.2,0.18369280099868773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,2,power_law_1.2,0.2231679916381836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,2,power_law_1.2,0.28326399326324464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,2,power_law_1.2,0.325926399230957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,2,power_law_1.2,0.44472317695617675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,2,power_law_1.2,0.5967807769775391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,2,power_law_1.2,0.8825920104980469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,2,power_law_1.2,1.2060159683227538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,2,power_law_1.2,1.372332763671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,2,power_law_1.2,2.0125375747680665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,2,power_law_1.2,3.9919872283935547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,2,power_law_1.2,0.044915199279785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,2,power_law_1.2,0.04356479942798615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,2,power_law_1.2,0.04699519872665405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,2,power_law_1.2,0.05480960011482239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,1,balanced,0.05021866659323374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,1,balanced,0.056373332937558494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,2,power_law_1.2,0.060031998157501223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,1,balanced,0.06833066542943318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,2,power_law_1.2,0.08280959725379944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,1,balanced,0.09477866689364116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,2,power_law_1.2,0.08831999897956848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,1,balanced,0.1429333289464315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,2,power_law_1.2,0.09194239974021912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,1,balanced,0.13804266850153604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,2,power_law_1.2,0.09131519794464112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,1,balanced,0.13461333513259888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,1,balanced,0.13726400335629782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,2,power_law_1.2,0.09265919923782348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,1,balanced,0.13662933309872946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,2,power_law_1.2,0.09484800100326538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,1,balanced,0.046282668908437095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,1,balanced,0.13821867108345032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,2,power_law_1.2,0.09852160215377807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,1,balanced,0.05157333115736643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,1,balanced,0.13793067137400308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,2,power_law_1.2,0.09797120094299316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,1,balanced,0.060517330964406334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,1,balanced,0.14136000474294028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,2,power_law_1.2,0.10257920026779174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,1,balanced,0.07913066446781158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,1,balanced,0.11223999659220378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,1,balanced,0.14127467075983682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,2,power_law_1.2,0.11010559797286987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,1,balanced,0.11558933059374492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,1,balanced,0.14517866571744284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,2,power_law_1.2,0.111244797706604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,1,balanced,0.11877333124478658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,2,power_law_1.2,0.12083200216293336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,1,balanced,0.15230933825174967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,1,balanced,0.11621333161989848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,2,power_law_1.2,0.1356608033180237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,1,balanced,0.1172213355700175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,1,balanced,0.15356799960136414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,1,balanced,0.12075733145078023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,2,power_law_1.2,0.15436160564422607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,1,balanced,0.16108266512552896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,1,balanced,0.12006400028864543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,2,power_law_1.2,0.19735679626464844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,1,balanced,0.12588799993197122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,2,power_law_1.2,0.22527999877929689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,1,balanced,0.17941866318384805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,1,balanced,0.12567466497421265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,2,power_law_1.2,0.29212160110473634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,1,balanced,0.19735999902089438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,1,balanced,0.13095466295878092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,2,power_law_1.2,0.3634624004364014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,1,balanced,0.14058666427930197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,2,power_law_1.2,0.5154176235198975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,1,balanced,0.14333333571751913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,1,balanced,0.23699732621510824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,2,power_law_1.2,0.7051775932312012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,1,balanced,0.1516746679941813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,2,power_law_1.2,0.9715583801269532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,1,balanced,0.17587733268737793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,1,balanced,0.2730240027109782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,1,balanced,0.1958400011062622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,1,balanced,0.4026133219401042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,1,balanced,0.23985600471496582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,1,balanced,0.281877338886261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,1,balanced,0.4777760108311971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,1,balanced,0.4294133186340332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,1,balanced,0.5189546744028727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,1,balanced,0.6714826424916586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,1,balanced,0.7467146714528402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,1,balanced,0.8532586892445883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,1,balanced,0.9714986483256022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,1,balanced,1.2541013558705647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,1,balanced,1.426143964131673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,1,power_law_1.01,0.046412798762321475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,1,balanced,1.863200028737386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,1,power_law_1.01,0.05258240103721619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,1,balanced,1.6335733731587727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,1,power_law_1.01,0.06287360191345215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,1,power_law_1.01,0.07888640165328979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,1,power_law_1.01,0.09017599821090698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,1,power_law_1.01,0.12613120079040527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,1,power_law_1.01,0.1252351999282837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,1,power_law_1.01,0.1278720021247864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,1,power_law_1.01,0.12728320360183715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,1,power_law_1.01,0.13207039833068848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,1,power_law_1.01,0.1302847981452942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,1,power_law_1.01,0.1368127942085266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,1,power_law_1.01,0.1385856032371521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,1,power_law_1.01,0.14369280338287355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,1,balanced,2.322362740834554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,1,power_law_1.01,0.14926719665527344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,1,balanced,2.0318506558736167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,1,power_law_1.01,0.15159039497375487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,1,power_law_1.01,0.16090879440307618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,1,power_law_1.01,0.1903167963027954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,1,power_law_1.01,0.2162623882293701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,1,power_law_1.01,0.26312320232391356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,1,power_law_1.01,0.305459189414978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,1,power_law_1.01,0.40778241157531736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,1,power_law_1.01,0.5051904201507569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,1,power_law_1.01,0.6997951984405517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,1,power_law_1.01,0.8834752082824707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,1,power_law_1.01,1.2878527641296387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,1,power_law_1.01,1.6756351470947266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,1,power_law_1.01,2.0743167877197264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,1,power_law_1.01,3.286073684692383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,1,power_law_1.01,6.547705841064453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,1,balanced,3.650501251220703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,1,balanced,3.2115678787231445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,1,balanced,7.239824295043945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,2,power_law_1.2,1.3606719970703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,2,power_law_1.2,1.6387456893920898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,2,power_law_1.2,2.6998336791992186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,2,power_law_1.2,5.254687881469726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,1,balanced,6.358874638875325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,1,power_law_1.01,0.04339199960231781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,1,power_law_1.01,0.04686079919338226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,1,power_law_1.01,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,1,power_law_1.01,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,1,power_law_1.01,0.077920001745224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,1,power_law_1.01,0.10316159725189208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,1,power_law_1.01,0.1066815972328186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,1,power_law_1.01,0.10869760513305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,1,power_law_1.01,0.11107200384140015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,1,power_law_1.01,0.11466879844665527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,1,power_law_1.01,0.11518080234527588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,1,power_law_1.01,0.1212224006652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,1,power_law_1.01,0.12388479709625244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,1,power_law_1.01,0.12756479978561402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,1,power_law_1.01,0.134060800075531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,1,power_law_1.01,0.140665602684021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,1,power_law_1.01,0.15155839920043945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,1,power_law_1.01,0.17984000444412232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,1,power_law_1.01,0.20633599758148194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,1,power_law_1.01,0.2657919883728027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,1,power_law_1.01,0.31359999179840087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,1,power_law_1.01,0.4289408206939697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,1,power_law_1.01,0.5407743930816651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,1,power_law_1.01,0.7662079811096192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,1,power_law_1.01,0.9809023857116699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,1,power_law_1.01,1.435814380645752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,1,power_law_1.01,1.882988739013672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,1,power_law_1.01,2.337094306945801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,1,power_law_1.01,3.68287353515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,1,power_law_1.01,7.313190460205078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,1,power_law_1.2,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,1,power_law_1.2,0.05247359871864319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,1,power_law_1.2,0.06050559878349304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,1,power_law_1.2,0.0716543972492218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,1,power_law_1.2,0.04318079948425293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,1,power_law_1.2,0.08071680068969726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,1,power_law_1.2,0.04767360091209412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,1,power_law_1.2,0.12177280187606812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,1,power_law_1.2,0.05272960066795349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,1,power_law_1.2,0.12620160579681397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,1,power_law_1.2,0.06414080262184144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,1,power_law_1.2,0.1283519983291626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,1,power_law_1.2,0.07557759881019592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,1,power_law_1.2,0.12893439531326295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,1,power_law_1.2,0.10373760461807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,1,power_law_1.2,0.13031680583953859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,1,power_law_1.2,0.10840959548950195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,1,power_law_1.2,0.13058559894561766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,1,power_law_1.2,0.10952320098876953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,1,power_law_1.2,0.1368127942085266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,1,power_law_1.2,0.11190400123596192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,1,power_law_1.2,0.13976320028305053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,16,balanced,0.0909493366877238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,1,power_law_1.2,0.11493120193481446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,16,balanced,0.09351467092831929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,1,power_law_1.2,0.1427839994430542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,16,balanced,0.10239467024803162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,1,power_law_1.2,0.11663999557495117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,1,power_law_1.2,0.15037440061569213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,16,balanced,0.14753066500027975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,1,power_law_1.2,0.12117760181427002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,1,power_law_1.2,0.15471359491348266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,16,balanced,0.21941866477330527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,1,power_law_1.2,0.12458239793777466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,1,power_law_1.2,0.16551040410995482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,16,balanced,0.25775466362635296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,1,power_law_1.2,0.12895359992980956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,1,power_law_1.2,0.20032000541687012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,16,balanced,0.23376532395680746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,1,power_law_1.2,0.1351680040359497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,1,power_law_1.2,0.222707200050354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,16,balanced,0.2478613257408142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,1,power_law_1.2,0.14331519603729248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,16,balanced,0.2761920094490051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,1,power_law_1.2,0.26652159690856936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,16,balanced,0.2603306571642558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,1,power_law_1.2,0.15334399938583373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,1,power_law_1.2,0.3101056098937988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,16,balanced,0.24580266078313193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,1,power_law_1.2,0.18229759931564332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,1,power_law_1.2,0.41006717681884763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,16,balanced,0.2634986639022827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,1,power_law_1.2,0.2111680030822754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,1,power_law_1.2,0.504697608947754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,16,balanced,0.2696479956309001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,1,power_law_1.2,0.2673791885375977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,1,power_law_1.2,0.7049856185913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,16,balanced,0.24788800875345865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,1,power_law_1.2,0.3151999950408936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,1,power_law_1.2,0.8944895744323731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,1,power_law_1.2,0.4306943893432617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,16,balanced,0.25863999128341675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,1,power_law_1.2,1.290112018585205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,1,power_law_1.2,0.540115213394165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,16,balanced,0.2472106615702311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,1,power_law_1.2,1.6834047317504883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,1,power_law_1.2,0.7682752132415771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,16,balanced,0.2493226726849874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,1,power_law_1.2,2.1026432037353517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,1,power_law_1.2,0.984160041809082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,1,power_law_1.2,3.3158782958984374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,16,balanced,0.24315732717514038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,1,power_law_1.2,1.4417152404785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,1,power_law_1.2,1.8839168548583984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,16,balanced,0.2470453381538391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,1,power_law_1.2,6.56622085571289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,1,power_law_1.2,2.345030403137207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,16,balanced,0.24405866861343384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,1,power_law_1.2,3.6846336364746093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,1,power_law_1.2,7.343923187255859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,16,balanced,0.2539786696434021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,16,balanced,0.4214826822280884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,16,balanced,0.3790239890416463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,16,balanced,0.5353813171386719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,16,balanced,0.7067786852518717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,16,balanced,0.7655786673227946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,16,balanced,1.1059679985046387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,16,balanced,0.06436266501744588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,16,balanced,0.06564266482988994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,16,balanced,1.1209920247395833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,16,balanced,0.07899199922879536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,16,balanced,0.10494933525721233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,16,balanced,0.14325333635012308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,16,balanced,0.15151466925938925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,16,balanced,0.15734933813412985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,16,balanced,0.15500799814860025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,16,balanced,0.1604373355706533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,16,balanced,0.14799466729164124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,16,balanced,0.16449600458145142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,16,balanced,0.16369066635767618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,16,balanced,0.1614773372809092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,16,balanced,0.16529599825541177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,16,balanced,0.16794667641321817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,16,balanced,0.16665066281954447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,16,balanced,0.16100800037384033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,16,balanced,0.17067732413609824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,16,balanced,0.17926400899887085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,16,balanced,0.18764267365137735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,16,balanced,0.20541866620381674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,16,balanced,1.7940746943155925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,16,balanced,0.3102826674779256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,16,balanced,0.3054719964663188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,16,balanced,0.04057066639264425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,16,balanced,0.04105599969625473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,16,balanced,0.046629334489504494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,16,balanced,0.06036800146102905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,16,balanced,0.04931733508904775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,16,balanced,0.06324266890684764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,16,balanced,0.4357973337173462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,16,balanced,0.05885866781075796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,16,balanced,0.0751039981842041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,16,balanced,0.10840533177057902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,16,balanced,0.06705066561698914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,16,balanced,0.15346666177113852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,16,balanced,0.08799466490745544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,16,balanced,0.1660160024960836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,16,balanced,0.08378666639328003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,16,balanced,0.14774933457374573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,16,balanced,0.10672533512115479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,16,balanced,0.1493119994799296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,16,balanced,0.10657599568367004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,16,balanced,0.16367999712626138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,16,balanced,0.1142026682694753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,16,balanced,0.14846932888031006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,16,balanced,0.14085333546002707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,16,balanced,0.5641866525014242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,16,balanced,0.15451199809710184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,16,balanced,0.1309760014216105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,16,balanced,0.15970666209856668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,16,balanced,0.13649066289265951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,16,balanced,0.149509330590566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,16,balanced,0.15202666322390238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,16,balanced,0.16990399360656738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,16,balanced,0.14915733536084494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,16,balanced,0.16940265893936157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,16,balanced,0.14094932874043783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,16,balanced,0.16149866580963135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,16,balanced,0.1544319987297058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,16,balanced,0.16149866580963135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,16,balanced,0.18218666315078735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,16,balanced,0.17143466075261435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,16,balanced,0.1630293329556783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,16,balanced,0.6797173023223877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,16,balanced,0.21941866477330527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,16,balanced,0.1662560005982717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,16,balanced,0.2602399984995524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,16,balanced,0.1721973419189453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,16,balanced,0.32814933856328327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,16,balanced,0.2731626629829407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,16,balanced,0.39576534430185956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,16,balanced,3.5356852213541665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,16,balanced,0.27214399973551434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,16,balanced,0.9478240013122559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,16,balanced,0.590399980545044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,16,balanced,0.3868693510691325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,16,balanced,0.7440959612528483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,16,balanced,0.4938240051269531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,16,balanced,1.0991146564483643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,16,balanced,1.0400479634602864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,16,balanced,0.568341334660848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,16,balanced,1.4396692911783855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,16,balanced,0.7949279944101969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,16,balanced,1.7329546610514324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,16,balanced,1.6359413464864094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,16,balanced,0.8558239936828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,16,balanced,2.7967147827148438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,16,balanced,1.3331466515858967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,16,balanced,3.24291197458903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,16,power_law_1.01,0.21578240394592285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,16,power_law_1.01,0.23788158893585204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,16,power_law_1.01,0.11640959978103638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,16,balanced,5.4696000417073565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,16,power_law_1.01,0.16365439891815187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,16,power_law_1.01,0.1829568028450012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,16,power_law_1.01,0.21818239688873292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,16,power_law_1.01,0.21623680591583253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,16,balanced,2.638352076212565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,16,power_law_1.01,0.24072959423065185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,16,power_law_1.01,0.23958399295806884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,16,power_law_1.01,0.2495487928390503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,16,power_law_1.01,0.23736319541931153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,16,power_law_1.01,0.2351423978805542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,16,power_law_1.01,0.23376638889312745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,16,power_law_1.01,0.24352641105651857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,16,power_law_1.01,0.24449920654296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,16,power_law_1.01,0.24664320945739746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,16,power_law_1.01,0.25899519920349123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,16,power_law_1.01,0.29534718990325926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,16,power_law_1.01,0.3070784091949463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,16,power_law_1.01,0.3561984062194824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,16,power_law_1.01,0.4200767993927002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,16,power_law_1.01,0.54202880859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,16,power_law_1.01,0.6250751972198486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,16,power_law_1.01,0.8192704200744629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,16,power_law_1.01,1.011315155029297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,16,power_law_1.01,1.5156224250793457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,16,power_law_1.01,1.997222328186035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,16,power_law_1.01,2.270572853088379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,16,power_law_1.01,3.5471424102783202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,16,power_law_1.01,7.938899230957031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,16,power_law_1.01,0.13671040534973145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,16,power_law_1.01,0.15189759731292723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,16,power_law_1.01,0.08118399977684021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,16,power_law_1.01,0.10785280466079712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,16,power_law_1.01,0.15676159858703614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,16,power_law_1.01,0.11318399906158447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,16,power_law_1.01,0.15482879877090455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,16,power_law_1.01,0.1391808032989502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,16,power_law_1.01,0.08254719972610473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,16,power_law_1.01,0.15426559448242189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,16,power_law_1.01,0.10448000431060792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,16,power_law_1.01,0.1498751997947693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,16,power_law_1.01,0.11883519887924195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,16,power_law_1.01,0.14977279901504517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,16,power_law_1.01,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,16,power_law_1.01,0.14332159757614135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,16,power_law_1.01,0.1540735960006714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,16,power_law_1.01,0.13815679550170898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,16,power_law_1.01,0.05779839754104614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,16,power_law_1.01,0.15934720039367675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,16,power_law_1.01,0.14414080381393432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,16,power_law_1.01,0.04087679982185364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,16,power_law_1.01,0.15834879875183105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,16,power_law_1.01,0.1555840015411377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,16,power_law_1.01,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,16,power_law_1.01,0.15847040414810182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,16,power_law_1.01,0.04912639856338501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,16,power_law_1.01,0.15326080322265626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,16,power_law_1.01,0.06399359703063964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,16,power_law_1.01,0.16403839588165284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,16,power_law_1.01,0.0871616005897522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,16,power_law_1.01,0.16828160285949706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,16,power_law_1.01,0.0836031973361969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,16,power_law_1.01,0.16810879707336426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,16,power_law_1.01,0.09048320055007934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,16,power_law_1.01,0.1974720001220703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,16,power_law_1.01,0.11048959493637085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,16,power_law_1.01,0.10648959875106812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,16,power_law_1.01,0.20842878818511962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,16,power_law_1.01,0.11367039680480957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,16,power_law_1.01,0.26426880359649657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,16,power_law_1.01,0.1376512050628662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,16,power_law_1.01,0.28663680553436277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,16,power_law_1.01,0.13479679822921753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,16,power_law_1.01,0.3694591999053955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,16,power_law_1.01,0.12465920448303222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,16,power_law_1.01,0.4693568229675293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,16,power_law_1.01,0.13000960350036622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,16,power_law_1.01,0.6090303897857666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,16,power_law_1.01,0.12120319604873657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,16,power_law_1.01,0.754310417175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,16,power_law_1.01,0.11247999668121338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,16,power_law_1.01,1.1693568229675293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.01,0.12082560062408447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,16,power_law_1.01,1.3110912322998047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.01,0.139136004447937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,16,power_law_1.01,1.7494911193847655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.01,0.16714240312576295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,16,power_law_1.01,2.739104080200195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.01,0.22098560333251954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.01,0.2460479974746704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,16,power_law_1.01,5.117254257202148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.01,0.3699520111083984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.01,0.4558976173400879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.01,0.644927978515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.01,0.8000831604003906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,16,power_law_1.01,1.021280002593994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,16,power_law_1.2,0.2133120059967041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,16,power_law_1.01,1.6739328384399415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,16,power_law_1.2,0.11976959705352783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,16,power_law_1.2,0.1118656039237976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,16,power_law_1.01,3.629529571533203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,16,power_law_1.2,0.1479871988296509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,16,power_law_1.2,0.18044159412384034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,16,power_law_1.2,0.22371840476989746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,16,power_law_1.2,0.23017599582672119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,16,power_law_1.2,0.24979839324951172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,16,power_law_1.2,0.23626880645751952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,16,power_law_1.2,0.23774080276489257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,16,power_law_1.2,0.2423935890197754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,16,power_law_1.2,0.22764160633087158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,16,power_law_1.2,0.24753921031951903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,16,power_law_1.2,0.23498880863189697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,16,power_law_1.2,0.2415424108505249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,16,power_law_1.2,0.2631488084793091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,16,power_law_1.2,0.289574408531189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,16,power_law_1.2,0.3215552091598511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,16,power_law_1.2,0.33147521018981935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,16,power_law_1.2,0.3948735952377319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,16,power_law_1.2,0.47825279235839846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,16,power_law_1.2,0.6506175994873047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,16,power_law_1.2,0.7842304229736328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,16,power_law_1.2,1.1378623962402343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,16,power_law_1.2,1.3819968223571777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,16,power_law_1.2,1.8890239715576171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,16,power_law_1.2,2.206233596801758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,16,power_law_1.2,3.6429824829101562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,16,power_law_1.2,5.714988708496094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,16,power_law_1.2,12.997158813476563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,16,power_law_1.01,0.15560959577560424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,16,power_law_1.01,0.15516159534454346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,16,power_law_1.01,0.152729594707489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,16,power_law_1.01,0.15692800283432007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,16,power_law_1.01,0.1536064028739929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,16,power_law_1.01,0.15859839916229249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,16,power_law_1.01,0.16172159910202027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,16,power_law_1.2,0.15061119794845582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,16,power_law_1.01,0.16778240203857422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,16,power_law_1.2,0.08789119720458985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,16,power_law_1.01,0.18789119720458985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,16,power_law_1.2,0.08103039860725403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.01,0.20528640747070312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,16,power_law_1.2,0.09745919704437256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.01,0.24764161109924315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,16,power_law_1.2,0.15436160564422607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,16,power_law_1.2,0.10090240240097045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.01,0.2666304111480713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,16,power_law_1.2,0.10580480098724365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,16,power_law_1.2,0.14950400590896606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.01,0.3531071901321411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,16,power_law_1.2,0.0944703996181488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,16,power_law_1.2,0.14439040422439575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.01,0.400486421585083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,16,power_law_1.2,0.09275519847869873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,16,power_law_1.2,0.15029120445251465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.01,0.5196800231933594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,16,power_law_1.2,0.10824320316314698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,16,power_law_1.2,0.15459200143814086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.01,0.6840447902679443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,16,power_law_1.2,0.1447551965713501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,16,power_law_1.2,0.15480320453643798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,16,power_law_1.2,0.1414528012275696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.01,0.9513152122497559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,16,power_law_1.2,0.1593727946281433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,16,power_law_1.2,0.14610559940338136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.01,1.2759424209594727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,16,power_law_1.2,0.15811840295791627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,16,power_law_1.2,0.1484544038772583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,16,power_law_1.01,1.5365119934082032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,16,power_law_1.2,0.14935679435729982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,16,power_law_1.01,2.052729606628418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,16,power_law_1.2,0.16270079612731933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,16,power_law_1.2,0.1529088020324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,16,power_law_1.01,4.533164978027344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,16,power_law_1.2,0.1542464017868042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,16,power_law_1.2,0.15758719444274902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,16,power_law_1.2,0.15861120223999023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,16,power_law_1.2,0.16388479471206666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,16,power_law_1.2,0.1707327961921692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,16,power_law_1.2,0.19742720127105712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.2,0.206329607963562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.2,0.24983680248260498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.2,0.3087615966796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.2,0.37695999145507814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.2,0.47447681427001953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.2,0.6483071804046631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.2,0.781324815750122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.2,1.2429823875427246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.2,1.390924835205078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,16,power_law_1.2,1.9465856552124023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,16,power_law_1.2,3.573606491088867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,16,power_law_1.2,6.315212631225586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,16,power_law_1.2,0.04912639856338501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,16,power_law_1.2,0.038278400897979736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,16,power_law_1.2,0.04030719995498657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,16,power_law_1.2,0.04031359851360321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,16,power_law_1.2,0.053235197067260744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,16,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,16,power_law_1.2,0.08394240140914917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,16,power_law_1.2,0.08384640216827392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,16,power_law_1.2,0.07607679963111877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,16,power_law_1.2,0.1118783950805664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,16,power_law_1.2,0.11166080236434936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,16,power_law_1.2,0.11269760131835938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,16,power_law_1.2,0.13969279527664186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,16,power_law_1.2,0.1305408000946045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,16,power_law_1.2,0.1245695948600769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,16,power_law_1.2,0.1155135989189148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,16,power_law_1.2,0.12773120403289795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,16,power_law_1.2,0.12140799760818481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.2,0.1276800036430359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.2,0.14881279468536376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.2,0.17047040462493895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.2,0.2263871908187866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.2,0.30236799716949464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.2,0.4162112236022949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.2,0.504095983505249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.2,0.7671360015869141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.2,1.0973504066467286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,16,power_law_1.2,1.2040767669677734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,16,power_law_1.2,2.0790592193603517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,16,power_law_1.2,4.032710266113281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,8,balanced,0.062496001521746315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,8,balanced,0.07810133198897044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,8,balanced,0.0918880005677541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,8,balanced,0.14403733611106873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,8,balanced,0.2153173287709554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,8,balanced,0.23433067401250204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,8,balanced,0.2446133295694987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,8,balanced,0.2270560065905253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,8,balanced,0.2457866668701172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,8,balanced,0.22913599014282227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,8,balanced,0.22659200429916382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,8,balanced,0.23174399137496948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,8,balanced,0.2425546646118164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,8,balanced,0.23757332563400269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,8,balanced,0.25649066766103107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,8,balanced,0.23942933479944864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,8,balanced,0.23306133349736533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,8,balanced,0.24472000201543173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,8,balanced,0.23827733596165976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,16,power_law_1.2,0.15790079832077025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,8,balanced,0.24727465709050497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,16,power_law_1.2,0.15859839916229249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,16,power_law_1.2,0.16331520080566406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,8,balanced,0.26001065969467163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,8,balanced,0.053397332628568016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,16,power_law_1.2,0.17240959405899048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,8,balanced,0.05829866727193197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,16,power_law_1.2,0.172326397895813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,8,balanced,0.06945066650708516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,16,power_law_1.2,0.19685759544372558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,8,balanced,0.09723200400670369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,8,balanced,0.41075201829274494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,8,balanced,0.1334879994392395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,16,power_law_1.2,0.22533760070800782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,8,balanced,0.14684266845385233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,16,power_law_1.2,0.2601855993270874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,8,balanced,0.154858668645223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,16,power_law_1.2,0.31166720390319824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,8,balanced,0.14573333660761514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,8,balanced,0.38439468542734784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,16,power_law_1.2,0.41095681190490724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,8,balanced,0.15332800149917603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,8,balanced,0.1402453382809957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,16,power_law_1.2,0.4983424186706543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,8,balanced,0.1458453337351481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,16,power_law_1.2,0.7359104156494141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,8,balanced,0.1518346667289734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,16,power_law_1.2,0.9221055984497071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,8,balanced,0.1495573321978251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,8,balanced,0.5431093374888102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,16,power_law_1.2,1.3898752212524415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,8,balanced,0.14696000019709268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,8,balanced,0.15825600425402322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,16,power_law_1.2,1.7640384674072265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,8,balanced,0.15832533439000449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,16,power_law_1.2,1.8711679458618165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,8,balanced,0.1556373337904612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,16,power_law_1.2,3.9795265197753906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,8,balanced,0.7062986691792806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,8,balanced,0.1699840029080709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,16,power_law_1.2,7.056352233886718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,8,balanced,0.17576533555984497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,8,balanced,0.19433599710464478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,8,balanced,0.21286400159200033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,8,balanced,0.8155093193054199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,8,balanced,0.30662399530410767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,8,balanced,0.3280479907989502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,8,balanced,1.1539519627888997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,8,balanced,0.4602080186208089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,8,balanced,0.5889600118001302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,8,balanced,1.1910346349080403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,8,balanced,0.05666666726271311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,8,balanced,0.06363200147946675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,8,balanced,0.07221333185831706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,8,balanced,0.09708266456921895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,8,balanced,0.13798399766286215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,8,balanced,0.15001066525777182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,8,balanced,0.7531786759694418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,8,balanced,0.15320533514022827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,8,balanced,0.1516320010026296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,8,balanced,0.1458133359750112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,8,balanced,0.15176000197728476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,8,balanced,0.15000533064206442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,8,balanced,0.16665599743525186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,8,balanced,0.1446506679058075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,8,balanced,0.14727466305096945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,8,balanced,0.17045867443084717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,8,balanced,0.15337066849072775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,8,balanced,0.16491732994715372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,8,balanced,0.1662346621354421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,8,balanced,1.0212159951527913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,8,balanced,1.9298879305521648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,8,balanced,0.164383997519811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,8,balanced,0.17657599846522012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,8,balanced,0.18492267529169717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,8,balanced,0.2850186626116435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,8,balanced,0.29598933458328247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,8,balanced,1.162394682566325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,8,balanced,0.4094346761703491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,8,balanced,0.5372373263041178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,8,balanced,0.642906665802002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,8,balanced,1.8291893005371094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,8,balanced,3.7915306091308594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,8,balanced,0.8903199831644694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,8,balanced,0.026911998788515728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,8,balanced,0.028677334388097126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,8,balanced,0.032986665765444435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,8,balanced,0.044735997915267944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,8,balanced,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,8,balanced,0.06699199974536896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,8,balanced,0.07710933188597362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,8,balanced,0.07421866556008656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,8,balanced,0.07877333462238312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,8,balanced,0.10237333178520203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,8,balanced,0.09582933783531189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,8,balanced,0.9862666924794515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,8,balanced,0.10461866855621338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,8,balanced,0.13532267014185587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,8,balanced,0.13828266660372415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,8,balanced,0.1446453332901001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,8,balanced,0.14947733283042908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,8,balanced,0.14940800269444784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,8,balanced,0.15523733695348105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,8,balanced,0.16204800208409628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,8,balanced,0.186191995938619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,8,balanced,0.20838934183120728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,8,balanced,0.24964267015457153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,8,balanced,3.609999974568685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,8,balanced,1.536608060201009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,8,balanced,0.2840426762898763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,8,balanced,0.4251893361409505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,8,balanced,0.5093173185984293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,8,balanced,0.7421973546346029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,8,balanced,0.9705813725789388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,8,balanced,3.0245386759440103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,8,balanced,1.1374239921569824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,8,power_law_1.01,0.1500864028930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,8,power_law_1.01,0.15580159425735474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,8,power_law_1.01,0.09833599925041199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,8,power_law_1.01,0.1349120020866394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,8,power_law_1.01,0.15701760053634645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,8,power_law_1.01,0.1981503963470459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,8,power_law_1.01,0.22053120136260987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,8,power_law_1.01,0.22802560329437255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,8,power_law_1.01,0.2230015993118286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,8,power_law_1.01,0.22234880924224854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,8,power_law_1.01,0.2260672092437744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,8,power_law_1.01,0.22430078983306884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,8,power_law_1.01,0.23113598823547363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,8,power_law_1.01,0.22803199291229248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,8,power_law_1.01,0.23171839714050294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,8,balanced,1.8453973134358723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,8,power_law_1.01,0.23622400760650636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,8,power_law_1.01,0.2465536117553711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,8,power_law_1.01,0.27295360565185545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,8,power_law_1.01,0.28725759983062743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,8,power_law_1.01,0.34209280014038085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,8,power_law_1.01,0.3790463924407959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,8,power_law_1.01,0.4643199920654297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,8,power_law_1.01,0.5902976036071778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,8,power_law_1.01,0.7118783950805664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,8,power_law_1.01,0.8993087768554687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,8,power_law_1.01,1.3489855766296386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,8,power_law_1.01,1.6427263259887694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,8,power_law_1.01,1.9139263153076171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,8,power_law_1.01,2.797113609313965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,8,power_law_1.01,6.43912353515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,8,power_law_1.01,0.09435520172119141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,8,power_law_1.01,0.1235584020614624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,8,power_law_1.01,0.07393280267715455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,8,power_law_1.01,0.0948095977306366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,8,power_law_1.01,0.10032639503479004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,8,power_law_1.01,0.13422720432281493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,8,power_law_1.01,0.13420159816741944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,8,power_law_1.01,0.14110080003738404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,8,power_law_1.01,0.14127999544143677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,8,power_law_1.01,0.14076160192489623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,8,power_law_1.01,0.14559359550476075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,8,power_law_1.01,0.14410239458084106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,8,power_law_1.01,0.15100159645080566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,8,power_law_1.01,0.15470080375671386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,8,power_law_1.01,0.15705599784851074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,8,balanced,3.556309382120768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,8,power_law_1.01,0.15852160453796388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,8,power_law_1.01,0.16681599617004395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,8,power_law_1.01,0.18188799619674684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,8,power_law_1.01,0.19635839462280275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,8,power_law_1.01,0.23767681121826173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,8,power_law_1.01,0.28803839683532717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,8,power_law_1.01,0.3788543939590454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,8,power_law_1.01,0.4175871849060059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,8,power_law_1.01,0.5559616088867188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,8,power_law_1.01,0.7475840091705322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,8,power_law_1.01,1.0081088066101074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,8,power_law_1.01,1.418892765045166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,8,power_law_1.01,1.5547967910766602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,8,power_law_1.01,2.738636779785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,8,power_law_1.01,4.706995010375977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,8,power_law_1.01,0.10472320318222046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,8,power_law_1.01,0.12281600236892701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,8,power_law_1.01,0.07373440265655518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,8,power_law_1.01,0.09652479887008666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,8,power_law_1.01,0.1066431999206543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,8,power_law_1.01,0.14442880153656007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,8,power_law_1.01,0.14796160459518432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,8,power_law_1.01,0.14700160026550294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,8,power_law_1.01,0.14926719665527344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,8,power_law_1.01,0.15066879987716675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,8,power_law_1.01,0.15026559829711914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,8,power_law_1.01,0.15329279899597167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,8,power_law_1.01,0.15444480180740355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,8,power_law_1.01,0.15393919944763185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,8,power_law_1.01,0.15879039764404296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,8,power_law_1.01,0.16199040412902832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,8,power_law_1.01,0.166048002243042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,8,power_law_1.01,0.18246400356292725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.01,0.20005760192871094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.01,0.2339008092880249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.01,0.25384318828582764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.01,0.33094398975372313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.01,0.40503678321838377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.01,0.5233920097351075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.01,0.6753151893615723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.01,0.9015935897827149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.01,1.097094440460205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,8,power_law_1.01,1.4247615814208985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,8,power_law_1.01,2.260966491699219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,8,power_law_1.01,4.484364700317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,8,power_law_1.01,0.03187839984893799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,8,power_law_1.01,0.04987519979476929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,8,power_law_1.01,0.03249920010566711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,8,power_law_1.01,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,8,power_law_1.01,0.04344959855079651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,8,power_law_1.01,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,8,power_law_1.01,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,8,power_law_1.01,0.06568319797515869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,8,power_law_1.01,0.0622655987739563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,8,power_law_1.01,0.08451840281486511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,8,power_law_1.01,0.08107519745826722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,8,power_law_1.01,0.07978879809379577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,8,power_law_1.01,0.11239039897918701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,8,power_law_1.01,0.11402239799499511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,8,power_law_1.01,0.1122048020362854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,8,power_law_1.01,0.1317247986793518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,8,power_law_1.01,0.12448639869689941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,8,power_law_1.01,0.12199039459228515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.01,0.11652480363845825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.01,0.13174400329589844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,8,power_law_1.2,0.14254720211029054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.01,0.14263039827346802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,8,power_law_1.2,0.10128639936447144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.01,0.1798784017562866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,8,power_law_1.2,0.1027135968208313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.01,0.21443839073181153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,8,power_law_1.2,0.12550400495529174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.01,0.2803136110305786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,8,power_law_1.2,0.14300160408020018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.01,0.3665472030639648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,8,power_law_1.2,0.20638079643249513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,8,power_law_1.2,0.22023680210113525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.01,0.4989312171936035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,8,power_law_1.2,0.22656641006469727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.01,0.6329599857330322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,8,power_law_1.2,0.21966080665588378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,8,power_law_1.01,0.7666368007659912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,8,power_law_1.2,0.2234879970550537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,8,power_law_1.01,1.1992704391479492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,8,power_law_1.2,0.23255040645599365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,8,power_law_1.2,0.23575680255889891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,8,power_law_1.01,2.5397760391235353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,8,power_law_1.2,0.23466238975524903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,8,power_law_1.2,0.22683520317077638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,8,power_law_1.2,0.24773120880126953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,8,power_law_1.2,0.2510848045349121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,8,power_law_1.2,0.09735680222511292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,8,power_law_1.2,0.2579008102416992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,8,power_law_1.2,0.06912000179290771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,8,power_law_1.2,0.2814591884613037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,8,power_law_1.2,0.06970880031585694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,8,power_law_1.2,0.31553919315338136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,8,power_law_1.2,0.09142400026321411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,8,power_law_1.2,0.10340479612350464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,8,power_law_1.2,0.3533504009246826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,8,power_law_1.2,0.1389631986618042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,8,power_law_1.2,0.40709757804870605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,8,power_law_1.2,0.14163199663162232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,8,power_law_1.2,0.502400016784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,8,power_law_1.2,0.1387712001800537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,8,power_law_1.2,0.6371327877044678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,8,power_law_1.2,0.14525439739227294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,8,power_law_1.2,0.9006208419799805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,8,power_law_1.2,0.14209280014038086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,8,power_law_1.2,0.9764032363891602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,8,power_law_1.2,0.14365439414978026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,8,power_law_1.2,1.4881088256835937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,8,power_law_1.2,0.15267839431762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,8,power_law_1.2,1.7908416748046876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,8,power_law_1.2,0.15336960554122925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,8,power_law_1.2,0.15044480562210083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,8,power_law_1.2,2.262579154968262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,8,power_law_1.2,0.16520960330963136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,8,power_law_1.2,4.101945495605468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,8,power_law_1.2,0.16416000127792357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,8,power_law_1.2,0.17056000232696533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,8,power_law_1.2,7.6642303466796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,8,power_law_1.2,0.20064001083374022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,8,power_law_1.2,0.21476480960845948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,8,power_law_1.2,0.2637696027755737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,8,power_law_1.2,0.3088128089904785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,8,power_law_1.2,0.3876800060272217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,8,power_law_1.2,0.45651841163635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,8,power_law_1.2,0.6688000202178955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,8,power_law_1.2,0.8030464172363281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,8,power_law_1.2,1.1369088172912598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,8,power_law_1.2,1.492959976196289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,8,power_law_1.2,1.9612096786499023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,8,power_law_1.2,2.997670364379883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,8,power_law_1.2,7.057094573974609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,8,power_law_1.2,0.10551680326461792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,8,power_law_1.2,0.0830847978591919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,8,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,8,power_law_1.2,0.08805760145187377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,8,power_law_1.2,0.10099200010299683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,8,power_law_1.2,0.14277119636535646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,8,power_law_1.2,0.14060159921646118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,8,power_law_1.2,0.1486847996711731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,8,power_law_1.2,0.14791680574417115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,8,power_law_1.2,0.14893440008163453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,8,power_law_1.2,0.14676480293273925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,8,power_law_1.2,0.14257279634475709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,8,power_law_1.2,0.1506880044937134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,8,power_law_1.2,0.15286400318145751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,8,power_law_1.2,0.1615679979324341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,8,power_law_1.2,0.1642303943634033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,8,power_law_1.2,0.17689599990844726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,8,power_law_1.2,0.19376640319824218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.2,0.21502079963684081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.2,0.2536895990371704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.2,0.29052159786224363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.2,0.34620161056518556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.2,0.4371903896331787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.2,0.5416319847106934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.2,0.7172607898712158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.2,1.0275327682495117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.2,1.3758336067199708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,8,power_law_1.2,1.7779584884643556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,8,power_law_1.2,2.6719423294067384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,8,power_law_1.2,5.671001434326172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,8,power_law_1.2,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,8,power_law_1.2,0.02587519884109497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,8,power_law_1.2,0.028672000765800475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,8,power_law_1.2,0.03813120126724243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,8,power_law_1.2,0.04134399890899658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,8,power_law_1.2,0.05507199764251709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,8,power_law_1.2,0.06258559823036194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,8,power_law_1.2,0.061536002159118655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,8,power_law_1.2,0.06424319744110107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,8,power_law_1.2,0.08218880295753479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,8,power_law_1.2,0.08167679905891419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,8,power_law_1.2,0.08086400032043457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,8,power_law_1.2,0.11285120248794556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,8,power_law_1.2,0.11472640037536622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,8,power_law_1.2,0.1162943959236145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,8,power_law_1.2,0.13551360368728638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,8,power_law_1.2,0.11794559955596924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,8,power_law_1.2,0.12078720331192017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.2,0.12131199836730958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.2,0.1280832052230835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.2,0.14762879610061647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,4,balanced,0.06055466830730438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.2,0.17615360021591187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,4,balanced,0.07567466795444489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.2,0.23756799697875977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,4,balanced,0.0942133367061615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,4,balanced,0.14136000474294028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.2,0.30305280685424807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,4,balanced,0.21100266774495444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.2,0.3931135892868042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,4,balanced,0.23380800088246664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.2,0.5599167823791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,4,balanced,0.2215786576271057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,4,balanced,0.22096532583236694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.2,0.7078720092773437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,4,balanced,0.2193546692530314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,8,power_law_1.2,0.8926464080810547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,4,balanced,0.21681066354115805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,4,balanced,0.2210879921913147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,8,power_law_1.2,1.4259455680847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,4,balanced,0.21434666713078818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,4,balanced,0.22574400901794434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,8,power_law_1.2,2.8976383209228516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,4,balanced,0.052570665876070656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,4,balanced,0.22403200467427573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,4,balanced,0.05762133498986562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,4,balanced,0.23563732703526816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,4,balanced,0.06824000179767609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,4,balanced,0.10466133554776509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,4,balanced,0.23720532655715942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,4,balanced,0.13492266337076822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,4,balanced,0.22797866662343344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,4,balanced,0.14563199877738953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,4,balanced,0.15169066190719604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,4,balanced,0.2430773377418518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,4,balanced,0.1437333325544993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,4,balanced,0.14893333117167154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,4,balanced,0.24563199281692505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,4,balanced,0.14630400141080221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,4,balanced,0.14149866501490274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,4,balanced,0.2637760043144226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,4,balanced,0.14789866407712302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,4,balanced,0.14869866768519083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,4,balanced,0.2853813370068868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,4,balanced,0.1497866710027059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,4,balanced,0.15586666266123453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,4,balanced,0.15745066603024802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,4,balanced,0.4219786723454793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,4,balanced,0.16379732886950174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,4,balanced,0.18025600910186768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,4,balanced,0.19056000312169394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,4,balanced,0.42477333545684814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,4,balanced,0.21878933906555176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,4,balanced,0.24339733521143594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,4,balanced,0.6011093457539877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,4,balanced,0.3412586847941081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,4,balanced,0.38302934169769287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,4,balanced,0.7751626968383789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,4,balanced,0.5373919804890951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,4,balanced,0.9385759830474854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,4,balanced,0.6911413669586182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,4,balanced,1.2812213102976482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,4,balanced,0.9130667050679525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,4,balanced,0.06249066690603892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,4,balanced,0.0674773355325063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,4,balanced,0.07815466821193695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,4,balanced,0.11174933115641277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,4,balanced,0.15618667006492615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,4,balanced,0.1795626680056254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,4,balanced,0.16850666205088297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,4,balanced,0.1622986694176992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,4,balanced,0.1673439939816793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,4,balanced,0.16457600394884744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,4,balanced,0.16296533743540445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,4,balanced,0.1613759994506836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,4,balanced,0.1672053337097168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,4,balanced,1.2154186566670735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,4,balanced,1.4363306363423665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,4,balanced,0.16313067078590393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,4,balanced,0.1693440079689026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,4,balanced,0.17072000106175741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,4,balanced,0.18239466349283853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,4,balanced,0.18352532386779785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,4,balanced,0.1964213252067566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,4,balanced,0.2099413275718689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,4,balanced,0.23146667083104452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,4,balanced,1.4288053512573242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,4,balanced,0.3511626720428467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,4,balanced,0.37468798955281574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,4,balanced,2.2925972938537598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,4,balanced,0.5216159820556641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,4,balanced,0.02566933383544286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,4,balanced,0.026752000053723652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,4,balanced,0.028575999041398365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,4,balanced,0.042026668787002563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,4,balanced,0.0580320010582606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,4,balanced,0.06923733154932658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,4,balanced,0.06970133384068807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,4,balanced,0.07272000114123027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,4,balanced,0.6697866916656494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,4,balanced,0.07340799768765767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,4,balanced,0.07734400033950806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,4,balanced,0.07546666761239369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,4,balanced,0.08122666676839192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,4,balanced,0.0990773340066274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,4,balanced,0.11086933811505635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,4,balanced,0.10688533385594685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,4,balanced,0.1415893336137136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,4,balanced,2.2494613329569497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,4,balanced,0.877994696299235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,4,balanced,0.15096533298492432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,4,balanced,0.16173866391181946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,4,balanced,0.16756266355514526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,4,balanced,0.18427733580271402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,4,balanced,0.19671465953191122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,4,balanced,0.221343994140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,4,balanced,1.1885653336842854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,4,balanced,0.2424586613972982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,4,balanced,0.3797333240509033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,4,balanced,1.3860054016113281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,4,balanced,0.4369279940923055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,4,balanced,4.531903902689616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,4,balanced,0.624181350072225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,4,balanced,4.449578603108724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,4,balanced,0.8174560070037842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,4,balanced,2.1615840593973794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,4,balanced,0.9368213017781576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,4,balanced,1.5190080006917317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,4,balanced,4.253114700317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,4,balanced,2.9133227666219077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,4,power_law_1.01,0.09605119824409485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,4,power_law_1.01,0.11887999773025512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,4,power_law_1.01,0.08878719806671143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,4,power_law_1.01,0.11236480474472046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,4,power_law_1.01,0.14744319915771484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,4,power_law_1.01,0.18897279500961303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,4,power_law_1.01,0.2027008056640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,4,power_law_1.01,0.2103679895401001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,4,power_law_1.01,0.21159679889678956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,4,power_law_1.01,0.21840639114379884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,4,power_law_1.01,0.21884799003601074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,4,power_law_1.01,0.07352960109710693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,4,power_law_1.01,0.21860480308532715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,4,power_law_1.01,0.08617600202560424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,4,power_law_1.01,0.22425599098205568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,4,power_law_1.01,0.06624000072479248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,4,power_law_1.01,0.22587521076202394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,4,power_law_1.01,0.08772479891777038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,4,power_law_1.01,0.22604799270629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,4,power_law_1.01,0.09838079810142517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,4,power_law_1.01,0.23082880973815917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,4,power_law_1.01,0.12865920066833497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,4,power_law_1.01,0.24145278930664063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,4,power_law_1.01,0.1360383987426758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,4,power_law_1.01,0.25477759838104247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,4,power_law_1.01,0.13375359773635864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,4,power_law_1.01,0.2699584007263184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,4,power_law_1.01,0.14410879611968994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,4,power_law_1.01,0.31940479278564454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,4,power_law_1.01,0.13957760334014893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,4,power_law_1.01,0.3678335905075073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,4,power_law_1.01,0.14648959636688233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,4,power_law_1.01,0.43753600120544434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,4,power_law_1.01,0.14575999975204468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,4,power_law_1.01,0.5551231861114502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,4,power_law_1.01,0.14736000299453736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,4,power_law_1.01,0.6931839942932129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,4,power_law_1.01,0.14816639423370362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,4,power_law_1.01,0.8970047950744628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,4,power_law_1.01,0.1603775978088379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,4,power_law_1.01,1.1839615821838378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,4,power_law_1.01,0.1628864049911499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,4,power_law_1.01,1.525216007232666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,4,power_law_1.01,0.17091840505599976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,4,power_law_1.01,1.9235071182250976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,4,power_law_1.01,0.1958400011062622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,4,power_law_1.01,3.034764862060547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,4,power_law_1.01,0.21301119327545165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,4,power_law_1.01,0.25287039279937745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,4,power_law_1.01,6.121247863769531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,4,power_law_1.01,0.2894848108291626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,4,power_law_1.01,0.3718208074569702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,4,power_law_1.01,0.4462463855743408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,4,power_law_1.01,0.6024960041046142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,4,power_law_1.01,0.7921472072601319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,4,power_law_1.01,1.143616008758545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,4,power_law_1.01,1.4246272087097167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,4,power_law_1.01,1.795782470703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,4,power_law_1.01,2.546131134033203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,4,power_law_1.01,5.300249481201172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,4,power_law_1.01,0.08190720081329346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,4,power_law_1.01,0.10090880393981934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,4,power_law_1.01,0.07406719923019409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,4,power_law_1.01,0.10287359952926636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,4,power_law_1.01,0.1120576024055481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,4,power_law_1.01,0.14952319860458374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,4,power_law_1.01,0.15017600059509278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,4,power_law_1.01,0.14981119632720946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,4,power_law_1.01,0.15754239559173583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,4,power_law_1.01,0.15427199602127076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,4,power_law_1.01,0.15438719987869262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,4,power_law_1.01,0.16013439893722534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,4,power_law_1.01,0.1562880039215088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,4,power_law_1.01,0.16644480228424072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,4,power_law_1.01,0.17488640546798706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,4,power_law_1.01,0.17358720302581787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,4,power_law_1.01,0.18180480003356933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,4,power_law_1.01,0.20432639122009277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.01,0.2187135934829712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.01,0.2579648017883301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.01,0.28659839630126954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.01,0.3743743896484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.01,0.4549568176269531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.01,0.6070208072662353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,4,power_law_1.01,0.02332800030708313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.01,0.7644351959228516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,4,power_law_1.01,0.03052160143852234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.01,1.0956671714782715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,4,power_law_1.01,0.027603200078010558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.01,1.4232383728027345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,4,power_law_1.01,0.036134400963783266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,4,power_law_1.01,1.6712831497192382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,4,power_law_1.01,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,4,power_law_1.01,2.656729507446289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,4,power_law_1.01,0.05496320128440857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,4,power_law_1.01,0.056435197591781616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,4,power_law_1.01,5.179411315917969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,4,power_law_1.01,0.06024320125579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,4,power_law_1.01,0.06120319962501526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,4,power_law_1.01,0.06544640064239501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,4,power_law_1.01,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,4,power_law_1.01,0.06776959896087646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,4,power_law_1.01,0.08329600095748901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,4,power_law_1.01,0.07526400089263915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,4,power_law_1.01,0.07841920256614685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,4,power_law_1.01,0.11575679779052735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,4,power_law_1.01,0.12508800029754638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,4,power_law_1.01,0.12565120458602905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.01,0.12325760126113891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.01,0.13500159978866577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.01,0.149126398563385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.01,0.17607040405273439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.01,0.2059648036956787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.01,0.2707456111907959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.01,0.33425920009613036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.01,0.45708799362182617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.01,0.5978623867034912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,4,power_law_1.01,0.7309951782226562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,4,power_law_1.01,1.1416255950927734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,4,power_law_1.01,2.236672019958496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,4,power_law_1.2,0.09386240243911743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,4,power_law_1.2,0.07954559922218322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,4,power_law_1.2,0.09027199745178223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,4,power_law_1.2,0.10347520112991333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,4,power_law_1.2,0.1310464024543762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,4,power_law_1.2,0.19300479888916017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,4,power_law_1.2,0.21281280517578124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,4,power_law_1.2,0.2173504114151001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,4,power_law_1.2,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,4,power_law_1.2,0.21337599754333497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,4,power_law_1.2,0.06256639957427979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,4,power_law_1.2,0.2188096046447754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,4,power_law_1.2,0.0703935980796814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,4,power_law_1.2,0.22442879676818847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,4,power_law_1.2,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,4,power_law_1.2,0.2112895965576172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,4,power_law_1.2,0.0937279999256134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,4,power_law_1.2,0.12830079793930055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,4,power_law_1.2,0.21519360542297364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,4,power_law_1.2,0.13100800514221192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,4,power_law_1.2,0.22072958946228027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,4,power_law_1.2,0.1380735993385315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,4,power_law_1.2,0.2304896116256714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,4,power_law_1.2,0.14108799695968627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,4,power_law_1.2,0.23516159057617186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,4,power_law_1.2,0.142848002910614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,4,power_law_1.2,0.24897921085357666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,4,power_law_1.2,0.145196795463562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,4,power_law_1.2,0.2803328037261963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,4,power_law_1.2,0.15050239562988282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,4,power_law_1.2,0.3022655963897705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,4,power_law_1.2,0.14639999866485595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,4,power_law_1.2,0.34177920818328855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,4,power_law_1.2,0.3850303888320923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,4,power_law_1.2,0.1522752046585083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,4,power_law_1.2,0.16244479417800903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,4,power_law_1.2,0.4470848083496094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,4,power_law_1.2,0.16807039976119995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,4,power_law_1.2,0.5889664173126221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,4,power_law_1.2,0.1760640025138855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,4,power_law_1.2,0.8479807853698731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,4,power_law_1.2,0.190937602519989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,4,power_law_1.2,0.964748764038086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,4,power_law_1.2,0.21989760398864747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,4,power_law_1.2,1.447481632232666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,4,power_law_1.2,0.26645760536193847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,4,power_law_1.2,1.6923839569091796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,4,power_law_1.2,0.3087807893753052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,4,power_law_1.2,2.0923648834228517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,4,power_law_1.2,0.37657599449157714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,4,power_law_1.2,3.634105682373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,4,power_law_1.2,0.4652416229248047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,4,power_law_1.2,0.620249605178833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,4,power_law_1.2,6.885273742675781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,4,power_law_1.2,0.8220479965209961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,4,power_law_1.2,1.1741696357727052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,4,power_law_1.2,1.5569151878356933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,4,power_law_1.2,2.003865623474121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,4,power_law_1.2,2.9219839096069338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,4,power_law_1.2,6.114720153808594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,4,power_law_1.2,0.08135039806365967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,4,power_law_1.2,0.06828160285949707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,4,power_law_1.2,0.07769600152969361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,4,power_law_1.2,0.09068800210952759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,4,power_law_1.2,0.10920319557189942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,4,power_law_1.2,0.14769279956817627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,4,power_law_1.2,0.16069120168685913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,4,power_law_1.2,0.1535423994064331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,4,power_law_1.2,0.16293120384216309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,4,power_law_1.2,0.1584447979927063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,4,power_law_1.2,0.1606528043746948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,4,power_law_1.2,0.16600960493087769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,4,power_law_1.2,0.17059839963912965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,4,power_law_1.2,0.16900479793548584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,4,power_law_1.2,0.17550079822540282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,4,power_law_1.2,0.18163199424743653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,4,power_law_1.2,0.18709759712219237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,4,power_law_1.2,0.21016960144042968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.2,0.2242432117462158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.2,0.2728512048721313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.2,0.30899839401245116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.2,0.4069632053375244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.2,0.4748095989227295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.2,0.654150390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,4,power_law_1.2,0.02343039959669113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.2,0.8361663818359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,4,power_law_1.2,0.025241601467132568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.2,1.3507455825805663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,4,power_law_1.2,0.02961919903755188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.2,1.5991488456726075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,4,power_law_1.2,0.03368319869041443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,4,power_law_1.2,1.8964351654052733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,4,power_law_1.2,0.04089600145816803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,4,power_law_1.2,3.0910207748413088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,4,power_law_1.2,0.05672320127487183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,4,power_law_1.2,0.059552001953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,4,power_law_1.2,6.167251205444336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,4,power_law_1.2,0.06050559878349304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,4,power_law_1.2,0.06221439838409424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,4,power_law_1.2,0.06291840076446534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,4,power_law_1.2,0.0689087986946106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,4,power_law_1.2,0.07047039866447449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,4,power_law_1.2,0.08484479784965515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,4,power_law_1.2,0.07726719975471497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,4,power_law_1.2,0.07644799947738648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,4,power_law_1.2,0.121343994140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,4,power_law_1.2,0.13014400005340576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,4,power_law_1.2,0.12337919473648071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.2,0.12034560441970825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.2,0.1494271993637085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.2,0.15340160131454467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.2,0.17471359968185424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.2,0.2150399923324585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.2,0.2718591928482056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.2,0.3290112018585205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.2,0.48127360343933107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.2,0.6359424114227294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,4,power_law_1.2,0.7841023921966552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,4,power_law_1.2,1.123583984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,4,power_law_1.2,2.3787071228027346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,2,balanced,0.06433600187301636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,2,balanced,0.07814933359622955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,2,balanced,0.056703999638557434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,2,balanced,0.09451733032862346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,2,balanced,0.06486933430035909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,2,balanced,0.14195733269055685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,2,balanced,0.07618666688601176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,2,balanced,0.21019200483957926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,2,balanced,0.11177066961924235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,2,balanced,0.2270080049832662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,2,balanced,0.15873600045839945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,2,balanced,0.23056000471115112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,2,balanced,0.172325332959493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,2,balanced,0.22003199656804404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,2,balanced,0.1650826632976532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,2,balanced,0.2207253376642863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,2,balanced,0.16879467169443765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,2,balanced,0.22433600823084512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,2,balanced,0.1655946671962738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,2,balanced,0.21894933780034384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,2,balanced,0.1657813290754954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,2,balanced,0.1660266617933909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,2,balanced,0.219760000705719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,2,balanced,0.17040000359217325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,2,balanced,0.2239840030670166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,2,balanced,0.1678239901860555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,2,balanced,0.23202133178710938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,2,balanced,0.17804267009099325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,2,balanced,0.2323946754137675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,2,balanced,0.1816693345705668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,2,balanced,0.23710399866104126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,2,balanced,0.18524267276128134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,2,balanced,0.24272000789642334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,2,balanced,0.19562133153279623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,2,balanced,0.2656906644503276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,2,balanced,0.21974400679270426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,2,balanced,0.2756800055503845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,2,balanced,0.23600532611211142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,2,balanced,0.2770559986432393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,2,balanced,0.3124799927075704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,2,balanced,0.3124799927075704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,2,balanced,0.3415679931640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,2,balanced,0.4507519801457723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,2,balanced,0.4827253421147664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,2,balanced,0.5213653246561686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,2,balanced,0.5361066659291586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,2,balanced,0.7344000339508057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,2,balanced,0.7436373233795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,2,balanced,0.947429339090983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,2,balanced,0.9562239646911621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,2,power_law_1.01,0.07735679745674133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,2,power_law_1.01,0.08906880021095276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,2,balanced,1.2904693285624187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,2,power_law_1.01,0.08807039856910706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,2,power_law_1.01,0.10699520111083985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,2,power_law_1.01,0.14494719505310058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,2,power_law_1.01,0.19094400405883788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,2,power_law_1.01,0.2050368070602417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,2,power_law_1.01,0.2032320022583008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,2,power_law_1.01,0.21073920726776124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,2,power_law_1.01,0.21004159450531007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,2,power_law_1.01,0.22132480144500732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,2,balanced,1.2272533575693767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,2,power_law_1.01,0.22286720275878907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,2,balanced,1.7181065877278645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,2,power_law_1.01,0.22799360752105713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,2,power_law_1.01,0.22393600940704345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,2,power_law_1.01,0.23745920658111572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,2,power_law_1.01,0.24348158836364747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,2,power_law_1.01,0.24863998889923095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,2,power_law_1.01,0.2715264081954956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,2,power_law_1.01,0.29706239700317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,2,power_law_1.01,0.06392959952354431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,2,power_law_1.01,0.34486401081085205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,2,power_law_1.01,0.0741375982761383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,2,power_law_1.01,0.37767040729522705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,2,power_law_1.01,0.07368320226669312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,2,power_law_1.01,0.4937151908874512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,2,power_law_1.01,0.09191039800643921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,2,power_law_1.01,0.5981887817382813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,2,power_law_1.01,0.1126207947731018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,2,balanced,2.048821290334066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,2,power_law_1.01,0.8051520347595215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,2,power_law_1.01,0.14536960124969484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,2,balanced,1.6471412976582844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,2,power_law_1.01,0.9519488334655761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,2,power_law_1.01,0.15663360357284545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,2,power_law_1.01,1.3357503890991211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,2,power_law_1.01,0.15649280548095704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,2,power_law_1.01,1.7261312484741211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,2,power_law_1.01,0.1609984040260315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,2,power_law_1.01,2.122265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,2,power_law_1.01,0.1621888041496277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,2,power_law_1.01,0.16219520568847656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,2,power_law_1.01,3.233113479614258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,2,power_law_1.01,0.16890239715576172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,2,power_law_1.01,6.604672241210937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,2,power_law_1.01,0.16921600103378295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,2,power_law_1.01,0.17957119941711425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,2,power_law_1.01,0.18391040563583375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,2,power_law_1.01,0.18762240409851075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,2,power_law_1.01,0.20258560180664062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,2,power_law_1.01,0.22883200645446777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,2,power_law_1.01,0.25440640449523927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,2,power_law_1.01,0.29913599491119386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,2,balanced,3.250901222229004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,2,power_law_1.01,0.35552000999450684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,2,balanced,1.9111572901407878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,2,power_law_1.01,0.45221757888793945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,2,power_law_1.01,0.585203218460083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,2,power_law_1.01,0.7775872230529786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,2,power_law_1.01,0.9919679641723633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,2,power_law_1.01,1.3938624382019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,2,power_law_1.01,1.8759231567382812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,2,power_law_1.01,2.2682815551757813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,2,power_law_1.01,3.6244224548339843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,2,power_law_1.01,7.018931579589844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,2,balanced,3.0178346633911133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,2,balanced,6.452698389689128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,2,balanced,5.974570592244466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,2,power_law_1.2,0.07683839797973632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,2,power_law_1.2,0.07454079985618592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,2,power_law_1.2,0.08829439878463745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,2,power_law_1.2,0.10974719524383544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,2,power_law_1.2,0.13846399784088134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,2,power_law_1.2,0.18981120586395264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,2,power_law_1.2,0.2006848096847534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,2,power_law_1.2,0.2114880084991455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,2,power_law_1.2,0.21916160583496094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,2,power_law_1.2,0.21697919368743895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,2,power_law_1.2,0.2180543899536133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,2,power_law_1.2,0.2232127904891968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,2,power_law_1.2,0.2260672092437744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,2,power_law_1.2,0.23301761150360106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,2,power_law_1.2,0.2383359909057617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,2,power_law_1.2,0.23924479484558106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,2,power_law_1.2,0.06389120221138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,2,power_law_1.2,0.26060800552368163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,2,power_law_1.2,0.06401919722557067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,2,power_law_1.2,0.28120319843292235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,2,power_law_1.2,0.07141759991645813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,2,power_law_1.2,0.31239678859710696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,2,power_law_1.2,0.0918079972267151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,2,power_law_1.2,0.34150400161743166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,2,power_law_1.2,0.10274560451507568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,2,power_law_1.2,0.40976638793945314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,2,power_law_1.2,0.13917440176010132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,2,power_law_1.2,0.15294079780578612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,2,power_law_1.2,0.5154431819915771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,2,power_law_1.2,0.1601408004760742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,2,power_law_1.2,0.6266751766204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,2,power_law_1.2,0.16085760593414306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,2,power_law_1.2,0.8408767700195312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,2,power_law_1.2,0.16353919506072997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,2,power_law_1.2,1.0088128089904784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,2,power_law_1.2,0.16679680347442627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,2,power_law_1.2,1.4199359893798829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,2,power_law_1.2,0.1737920045852661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,2,power_law_1.2,1.83702392578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,2,power_law_1.2,0.17110400199890136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,2,power_law_1.2,2.200351905822754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,2,power_law_1.2,0.18136320114135743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,2,power_law_1.2,0.18703360557556153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,2,power_law_1.2,3.855955123901367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,2,power_law_1.2,0.18995840549468995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,2,power_law_1.2,6.797190093994141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,2,power_law_1.2,0.20472960472106932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,2,power_law_1.2,0.23011839389801025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,2,power_law_1.2,0.2612031936645508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,1,balanced,0.07099199791749318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,2,power_law_1.2,0.3069888114929199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,1,balanced,0.08467732866605122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,2,power_law_1.2,0.3625727891921997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,1,balanced,0.10851732889811198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,1,balanced,0.16285333037376404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,2,power_law_1.2,0.4574399948120117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,1,balanced,0.24934399127960205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,2,power_law_1.2,0.5871424198150634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,1,balanced,0.29418667157491046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,2,power_law_1.2,0.7489471912384034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,1,balanced,0.2839999993642171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,1,balanced,0.2835093339284261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,2,power_law_1.2,1.0811327934265136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,1,balanced,0.2851360042889913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,2,power_law_1.2,1.4439616203308105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,1,balanced,0.279146671295166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,1,balanced,0.2858346700668335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,2,power_law_1.2,1.8564863204956055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,1,balanced,0.28870399792989093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,2,power_law_1.2,2.3905792236328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,1,balanced,0.28916800022125244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,1,balanced,0.29581334193547565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,2,power_law_1.2,3.7221569061279296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,1,balanced,0.30163200696309406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,2,power_law_1.2,6.993011474609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,1,balanced,0.30612800518671673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,1,balanced,0.3158026734987895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,1,balanced,0.35288000106811523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,1,balanced,0.3767840067545573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,1,balanced,0.43421868483225506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,1,balanced,0.49021867911020917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,1,balanced,0.718010663986206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,1,balanced,0.8138666947682699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,1,balanced,1.158186674118042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,1,balanced,1.497104008992513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,1,balanced,0.06663999954859416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,1,balanced,0.07712533573309581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,1,balanced,0.09517866373062134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,1,balanced,1.9995147387186687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,1,balanced,0.13896000385284424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,1,balanced,0.20148799816767374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,1,balanced,0.23097066084543863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,1,balanced,0.22258667151133218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,1,balanced,0.2259733279546102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,1,balanced,0.22683733701705933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,1,balanced,0.22236265738805136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,1,balanced,0.22748267650604248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,1,balanced,0.22721600532531738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,1,balanced,0.23348265886306763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,1,balanced,0.23492266734441122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,1,balanced,0.24544000625610352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,1,balanced,2.6760854721069336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,1,balanced,0.2521013418833415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,1,balanced,0.26420799891153973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,1,balanced,0.3027199904123942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,1,balanced,0.32893866300582886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,1,balanced,0.3946559826533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,1,balanced,0.45159467061360675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,1,balanced,3.188096046447754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,1,balanced,0.676245371500651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,1,balanced,0.784997304280599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,1,balanced,1.1423412958780925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,1,balanced,1.4891093571980794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,1,power_law_1.01,0.06892160177230836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,1,balanced,5.056810696919759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,1,power_law_1.01,0.0806656002998352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,1,power_law_1.01,0.09735680222511292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,1,power_law_1.01,0.1257151961326599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,1,power_law_1.01,0.16830719709396363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,1,power_law_1.01,0.2329279899597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,1,balanced,2.0442399978637695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,1,power_law_1.01,0.2330944061279297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,1,power_law_1.01,0.24035840034484862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,1,power_law_1.01,0.24783999919891359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,1,power_law_1.01,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,1,power_law_1.01,0.2476736068725586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,1,power_law_1.01,0.25591039657592773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,1,power_law_1.01,0.0701632022857666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,1,power_law_1.01,0.0833407998085022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,1,power_law_1.01,0.26201601028442384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,1,power_law_1.01,0.10766719579696656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,1,power_law_1.01,0.25988481044769285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,1,power_law_1.01,0.1312831997871399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,1,power_law_1.01,0.27808640003204343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,1,power_law_1.01,0.18566399812698364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,1,power_law_1.01,0.27957758903503416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,1,balanced,2.729584058125814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,1,power_law_1.01,0.2969343900680542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,1,power_law_1.01,0.1918015956878662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,1,power_law_1.01,0.19148800373077393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,1,power_law_1.01,0.31578240394592283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,1,power_law_1.01,0.19711999893188475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,1,power_law_1.01,0.3568063974380493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,1,power_law_1.01,0.20013439655303955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,1,power_law_1.01,0.3855616092681885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,1,power_law_1.01,0.20830080509185792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,1,power_law_1.01,0.47069439888000486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,1,power_law_1.01,0.20655360221862792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,1,power_law_1.01,0.5387328147888184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,1,power_law_1.01,0.2170111894607544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,1,power_law_1.01,0.6902783870697021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,1,power_law_1.01,0.22130560874938965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,1,power_law_1.01,0.8423680305480957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,1,power_law_1.01,0.23399679660797118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,1,power_law_1.01,1.1431872367858886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,1,power_law_1.01,0.23878400325775145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,1,power_law_1.01,1.4476927757263183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,1,power_law_1.01,0.25864319801330565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,1,power_law_1.01,2.0649471282958984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,1,balanced,3.313722610473633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,1,power_law_1.01,0.3040895938873291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,1,power_law_1.01,2.68670711517334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,1,power_law_1.01,0.33566720485687257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,1,power_law_1.01,3.29881591796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,1,power_law_1.01,0.4213568210601807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,1,power_law_1.01,0.4945087909698486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,1,power_law_1.01,5.1735679626464846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,1,power_law_1.01,0.6557248115539551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,1,power_law_1.01,0.8113408088684082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,1,power_law_1.01,10.202047729492188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,1,power_law_1.01,1.1240127563476563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,1,balanced,10.056592305501303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,1,power_law_1.01,1.4377408027648926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,1,power_law_1.01,2.076019287109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,1,power_law_1.01,2.701433563232422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,1,power_law_1.01,3.3527553558349608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,1,power_law_1.01,5.294137573242187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,1,power_law_1.01,10.496633911132813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,1,balanced,5.233695983886719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,1,balanced,10.407855987548828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,1,power_law_1.2,0.069651198387146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,1,power_law_1.2,0.07854080200195312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,1,power_law_1.2,0.06219519972801209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,1,power_law_1.2,0.09536640048027038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,1,power_law_1.2,0.0701632022857666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,1,power_law_1.2,0.11883519887924195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,16,balanced,0.08532800277074178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,1,power_law_1.2,0.08490880131721497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,16,balanced,0.08610133330027263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,1,power_law_1.2,0.15601919889450072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,16,balanced,0.09571733077367146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,1,power_law_1.2,0.10428800582885742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,1,power_law_1.2,0.22692480087280273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,16,balanced,0.1290079951286316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,1,power_law_1.2,0.12668800354003906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,1,power_law_1.2,0.23994879722595214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,16,balanced,0.18685332934061685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,1,power_law_1.2,0.1875712037086487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,1,power_law_1.2,0.245580792427063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,16,balanced,0.33078400293986004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,16,balanced,0.45366934935251874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,1,power_law_1.2,0.18889600038528442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,1,power_law_1.2,0.2519680023193359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,16,balanced,0.46860265731811523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,1,power_law_1.2,0.19905279874801635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,1,power_law_1.2,0.2498431921005249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,16,balanced,0.4538826545079549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,1,power_law_1.2,0.20056960582733155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,1,power_law_1.2,0.25824000835418703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,16,balanced,0.47041066487630206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,1,power_law_1.2,0.2754688024520874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,1,power_law_1.2,0.20670719146728517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,16,balanced,0.4714346726735433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,1,power_law_1.2,0.2764928102493286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,1,power_law_1.2,0.2088831901550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,16,balanced,0.46136001745859784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,1,power_law_1.2,0.284115195274353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,16,balanced,0.46291200319925946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,1,power_law_1.2,0.2182080030441284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,16,balanced,0.47018667062123615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,1,power_law_1.2,0.301580810546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,1,power_law_1.2,0.2230976104736328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,1,power_law_1.2,0.23032960891723633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,16,balanced,0.4597333272298177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,1,power_law_1.2,0.24313600063323976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,16,balanced,0.4570986827214559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,1,power_law_1.2,0.2516160011291504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,16,balanced,0.4785279830296834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,1,power_law_1.2,0.2656064033508301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,16,balanced,0.45762133598327637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,1,power_law_1.2,0.3114943981170654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,16,balanced,0.4764426549275716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,1,power_law_1.2,0.34082560539245604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,1,power_law_1.2,0.42924160957336427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,16,balanced,0.49219199021657306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,1,power_law_1.2,0.5062464237213135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,1,power_law_1.2,0.6648064136505127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,16,balanced,0.48284268379211426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,1,power_law_1.2,0.8163455963134766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,1,power_law_1.2,1.1279232025146484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,1,power_law_1.2,1.4475199699401855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,16,balanced,0.5032586654027303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,1,power_law_1.2,2.082624053955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,1,power_law_1.2,2.7147775650024415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,16,balanced,0.5247146685918173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,1,power_law_1.2,3.3702144622802734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,1,power_law_1.2,5.32487678527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,1,power_law_1.2,10.494195556640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,16,balanced,0.5910666783650717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,16,balanced,0.9194506804148356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,16,balanced,0.9035360018412272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,16,balanced,0.06479466458161671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,16,balanced,0.06855999926726024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,16,balanced,0.06912533442179362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,16,balanced,0.09194667140642802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,16,balanced,0.13294933239618936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,16,balanced,0.20760534207026163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,16,balanced,0.28064533074696857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,16,balanced,0.2649280031522115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,16,balanced,0.267797331015269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,16,balanced,1.2723146279652913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,16,balanced,0.2648106614748637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,16,balanced,0.27000532547632855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,16,balanced,0.27347199122111004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,16,balanced,0.26521599292755127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,16,balanced,0.268559992313385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,16,balanced,0.2797866662343343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,16,balanced,0.2624586621920268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,16,balanced,0.29530133803685504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,16,balanced,0.2792960007985433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,16,balanced,0.30130134026209515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,16,balanced,0.31683733065923053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,16,balanced,1.718015988667806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,16,balanced,0.33935999870300293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,16,balanced,0.3782133261362712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,16,balanced,0.41969064871470135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,16,balanced,0.5140959819157919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,16,balanced,0.7294560273488363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,16,balanced,2.4860266049702964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,16,balanced,0.8561600049336752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,16,balanced,1.151920000712077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,16,balanced,1.5053226153055828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,1,power_law_1.2,0.30264959335327146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,1,power_law_1.2,0.32094080448150636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,1,power_law_1.2,0.36094720363616944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,1,power_law_1.2,0.39271039962768556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,1,power_law_1.2,0.4722559928894043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,1,power_law_1.2,0.5495488166809082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,1,power_law_1.2,0.7019839763641358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,1,power_law_1.2,0.8491583824157715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,16,balanced,0.06488533318042755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,1,power_law_1.2,1.1479552268981934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,16,balanced,0.06698133548100789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,1,power_law_1.2,1.4526847839355468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,16,balanced,0.07665599882602692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,16,balanced,4.317178726196289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,16,balanced,0.10844799876213074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,1,power_law_1.2,2.0730623245239257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,16,balanced,0.14921067158381143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,1,power_law_1.2,2.688627243041992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,16,balanced,0.20390399297078451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,16,balanced,0.29339200258255005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,1,power_law_1.2,3.3137535095214843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,16,balanced,0.2640693386395772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,16,balanced,0.261189341545105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,1,power_law_1.2,5.160819244384766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,16,balanced,0.2560799916585286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,16,balanced,0.24515734116236368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,1,power_law_1.2,10.20487060546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,16,balanced,0.2655893365542094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,16,balanced,2.258863925933838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,16,balanced,0.2600213289260864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,16,balanced,0.25042132536570233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,16,balanced,0.2815093398094177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,16,balanced,0.26904000838597614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,16,balanced,0.26895999908447266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,16,balanced,0.2681120038032532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,16,balanced,0.2857813239097595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,16,balanced,0.29101866483688354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,16,balanced,0.31276800235112506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,16,balanced,0.3279413382212321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,16,balanced,0.35024531682332355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,16,balanced,0.41513065497080487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,16,balanced,0.6478293339411417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,16,balanced,4.124202728271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,16,balanced,0.708575963973999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,16,balanced,0.9726133346557617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,16,balanced,1.2870773474375408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,16,power_law_1.01,0.12485760450363159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,16,power_law_1.01,0.19466880559921265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,16,power_law_1.01,0.19366400241851806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,16,power_law_1.01,0.341644811630249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,16,power_law_1.01,0.25281920433044436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,16,power_law_1.01,0.46285438537597656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,16,power_law_1.01,0.10644479990005493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,16,power_law_1.01,0.13388799428939818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,16,power_law_1.01,0.11927039623260498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,16,power_law_1.01,0.20633599758148194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,16,power_law_1.01,0.16943999528884887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,16,power_law_1.01,0.26841599941253663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,16,power_law_1.01,0.18114559650421141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,16,power_law_1.01,0.29339520931243895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,16,power_law_1.01,0.2188800096511841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,16,power_law_1.01,0.35878400802612304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,16,power_law_1.01,0.2055295944213867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,16,power_law_1.01,0.335315203666687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,16,balanced,1.9236960411071777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,16,power_law_1.01,0.20427520275115968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,16,power_law_1.01,0.3536128044128418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,16,power_law_1.01,0.23081600666046143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,16,power_law_1.01,0.35493760108947753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,16,power_law_1.01,0.22486400604248047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,16,power_law_1.01,0.3814784049987793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,16,power_law_1.01,0.2362816095352173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,16,power_law_1.01,0.38375039100646974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,16,power_law_1.01,0.23086080551147461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,16,power_law_1.01,0.3940288066864014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,16,power_law_1.01,0.24085121154785155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,16,power_law_1.01,0.4075456142425537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,16,power_law_1.01,0.24970879554748535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,16,power_law_1.01,0.4142911911010742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,16,power_law_1.01,0.24160640239715575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,16,power_law_1.01,0.41559038162231443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,16,power_law_1.01,0.26742401123046877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,16,power_law_1.01,0.4260608196258545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,16,power_law_1.01,0.2940608024597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,16,power_law_1.01,0.4504511833190918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,16,power_law_1.01,0.3275007963180542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,16,power_law_1.01,0.4907519817352295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,16,power_law_1.01,0.36792960166931155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,16,power_law_1.01,0.5415552139282227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,16,power_law_1.01,0.47038722038269043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,16,power_law_1.01,0.6386688232421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,16,power_law_1.01,0.5547327995300293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,16,power_law_1.01,0.7503744125366211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,16,power_law_1.01,0.6757952213287354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,16,power_law_1.01,1.0148223876953124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,16,power_law_1.01,0.8793472290039063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,16,power_law_1.01,1.1774399757385254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,16,power_law_1.01,1.1539008140563964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,16,power_law_1.01,1.5346688270568847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,16,power_law_1.01,1.4423871994018556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,16,power_law_1.01,1.894278335571289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,16,power_law_1.01,1.8536384582519532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,16,power_law_1.01,2.4206527709960937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,16,power_law_1.01,2.752729606628418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,16,power_law_1.01,3.5625919342041015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,16,power_law_1.01,5.487494277954101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,16,power_law_1.01,6.68748779296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,16,power_law_1.01,0.13216639757156373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,16,power_law_1.01,0.22385919094085693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,16,power_law_1.01,0.2840127944946289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,16,power_law_1.01,0.11498240232467652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,16,power_law_1.01,0.1360640048980713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,16,balanced,3.428895950317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,16,power_law_1.01,0.18183679580688478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,16,power_law_1.01,0.19649280309677125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,16,power_law_1.01,0.21144959926605225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,16,power_law_1.01,0.2200000047683716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,16,power_law_1.01,0.21766400337219238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,16,power_law_1.01,0.23175039291381835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,16,power_law_1.01,0.23678719997406006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,16,power_law_1.01,0.22716801166534423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,16,power_law_1.01,0.24814081192016602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,16,power_law_1.01,0.2517568111419678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,16,power_law_1.01,0.24717440605163574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,16,power_law_1.01,0.24360320568084717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,16,power_law_1.01,0.2636735916137695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,16,power_law_1.01,0.2790719985961914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,16,power_law_1.01,0.30672640800476075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,16,power_law_1.01,0.3585855960845947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,16,power_law_1.01,0.4175871849060059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,16,power_law_1.01,0.49228801727294924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,16,power_law_1.01,0.6464704036712646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,16,power_law_1.01,0.7690624237060547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,16,power_law_1.01,1.0180031776428222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,16,power_law_1.01,1.2842368125915526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,16,power_law_1.01,1.6408575057983399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,16,power_law_1.01,2.469196891784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,16,power_law_1.01,4.774566268920898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,16,power_law_1.2,0.1978816032409668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,16,power_law_1.2,0.3382591962814331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,16,power_law_1.2,0.12691839933395385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,16,power_law_1.2,0.2670720100402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,16,power_law_1.2,0.19199999570846557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,16,power_law_1.2,0.15319039821624755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,16,power_law_1.2,0.1468287944793701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,16,power_law_1.2,0.19900799989700318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,16,power_law_1.2,0.10170880556106568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,16,power_law_1.2,0.24280319213867188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,16,power_law_1.2,0.11811200380325318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,16,power_law_1.2,0.2804352045059204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,16,power_law_1.2,0.15294079780578612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,16,power_law_1.2,0.1299008011817932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,16,power_law_1.2,0.32469758987426756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,16,power_law_1.2,0.17580800056457518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,16,power_law_1.2,0.20814080238342286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,16,power_law_1.2,0.34999039173126223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,16,power_law_1.2,0.212556791305542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,16,power_law_1.2,0.22641921043395996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,16,power_law_1.2,0.35288960933685304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,16,power_law_1.2,0.20655999183654786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,16,power_law_1.2,0.10805120468139648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,16,power_law_1.2,0.3718015909194946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,16,power_law_1.2,0.13528319597244262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,16,power_law_1.2,0.2149183988571167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,16,power_law_1.2,0.36919679641723635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,16,power_law_1.2,0.17596800327301027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,16,power_law_1.2,0.21955840587615966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,16,power_law_1.2,0.3507776021957397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,16,power_law_1.2,0.18504960536956788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,16,power_law_1.2,0.21650559902191163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,16,power_law_1.2,0.357094407081604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,16,power_law_1.2,0.22305920124053955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,16,power_law_1.2,0.24089601039886474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,16,power_law_1.2,0.3817536115646362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,16,power_law_1.2,0.21300480365753174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,16,power_law_1.2,0.22799360752105713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,16,power_law_1.2,0.3920000076293945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,16,power_law_1.2,0.21639039516448974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,16,power_law_1.2,0.24165120124816894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,16,power_law_1.2,0.4052800178527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,16,power_law_1.2,0.2231231927871704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,16,power_law_1.2,0.25051519870758054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,16,power_law_1.2,0.4348608016967773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,16,power_law_1.2,0.22186880111694335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,16,power_law_1.2,0.25747840404510497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,16,power_law_1.2,0.47632641792297364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,16,power_law_1.2,0.22462079524993897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,16,power_law_1.2,0.29331839084625244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,16,power_law_1.2,0.234169602394104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,16,power_law_1.2,0.5314303874969483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,16,power_law_1.2,0.303219199180603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,16,power_law_1.2,0.26207358837127687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,16,power_law_1.2,0.5789120197296143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,16,power_law_1.2,0.3584000110626221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,16,power_law_1.2,0.2525887966156006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,16,power_law_1.2,0.7098944187164307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,16,power_law_1.2,0.39855360984802246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,16,power_law_1.2,0.26225919723510743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,16,power_law_1.2,0.8342399597167969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,16,power_law_1.2,0.472979211807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,16,power_law_1.2,0.27895679473876955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,16,power_law_1.2,0.6193215847015381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,16,power_law_1.2,1.0255423545837403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,16,power_law_1.2,0.29492480754852296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,16,power_law_1.2,1.3936191558837892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,16,power_law_1.2,0.7461696147918702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,16,power_law_1.2,0.3422784090042114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,16,power_law_1.2,0.9307392120361329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,16,power_law_1.2,1.7057151794433594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,16,power_law_1.2,0.3876735925674438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,16,power_law_1.2,1.2824447631835938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,16,power_law_1.2,2.051430320739746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,16,power_law_1.2,0.4669312000274658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,16,power_law_1.2,1.5594816207885742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,16,power_law_1.2,2.732697677612305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,16,power_law_1.2,0.5374591827392579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,16,power_law_1.2,1.975334358215332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,16,power_law_1.2,4.0691070556640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,16,power_law_1.2,0.7345791816711426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,16,power_law_1.2,3.088140869140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,16,power_law_1.2,0.8353983879089355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,16,power_law_1.2,8.345331573486328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,16,power_law_1.2,1.151692771911621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,16,power_law_1.2,6.449049377441407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,16,power_law_1.2,1.524403190612793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,16,power_law_1.2,1.7662208557128907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,16,power_law_1.2,3.0962240219116213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,16,power_law_1.2,5.760550308227539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,8,balanced,0.06800533334414165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,8,balanced,0.07986133297284444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,8,balanced,0.09365333120028178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,8,balanced,0.12117866675059001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,8,balanced,0.1861706574757894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,8,balanced,0.3062613407770793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,8,balanced,0.43301868438720703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,8,balanced,0.44305598735809326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,8,balanced,0.42953598499298096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,8,balanced,0.4344000021616618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,8,balanced,0.4325546820958455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,8,balanced,0.4537866512934367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,8,balanced,0.4463626543680827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,8,balanced,0.4406506617863973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,8,balanced,0.4479680061340332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,8,balanced,0.4463253418604533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,8,balanced,0.44678934415181476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,8,balanced,0.45375998814900714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,8,balanced,0.4723626772562663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,8,balanced,0.47273067633310956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,8,balanced,0.4972746769587199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,8,balanced,0.517904003461202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,8,balanced,0.545637329419454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,8,balanced,0.626309315363566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,8,balanced,0.9265279769897461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,8,balanced,0.9732960065205892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,8,balanced,1.3158453305562336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,8,balanced,0.054645334680875145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,8,balanced,0.0565280020236969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,8,balanced,1.7874080340067546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,8,balanced,0.06330133477846782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,8,balanced,0.06558933357397716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,8,balanced,0.07007466753323872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,8,balanced,0.07867733140786488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,8,balanced,0.09196799993515015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,8,balanced,0.10363733768463135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,8,balanced,0.124917338291804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,8,balanced,0.13833600282669067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,8,balanced,0.19710934162139893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,8,balanced,0.2148639957110087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,8,balanced,0.25730667511622113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,8,balanced,0.2847573359807332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,8,balanced,0.2474773327509562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,8,balanced,0.2722613414128621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,8,balanced,0.2651360034942627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,8,balanced,0.2565866708755493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,8,balanced,0.2562933365503947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,8,balanced,0.2460213303565979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,8,balanced,0.24908800919850668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,8,balanced,0.2539946635564168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,8,balanced,0.2512586712837219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,8,balanced,0.24632000923156738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,8,balanced,0.25226666529973346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,8,balanced,0.24906132618586221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,8,balanced,0.25045865774154663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,8,balanced,0.25948800643285114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,8,balanced,0.27292799949645996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,8,balanced,0.26107199986775714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,8,power_law_1.01,0.12554880380630493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,8,balanced,0.26941333214441937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,8,balanced,0.252895991007487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,8,power_law_1.01,0.19815679788589477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,8,balanced,0.27698665857315063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,8,balanced,0.26412800947825116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,8,power_law_1.01,0.30213758945465086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,8,balanced,0.28684266408284503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,8,balanced,0.2632213234901428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,8,power_law_1.01,0.13129600286483764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,8,balanced,0.3016480008761088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,8,balanced,0.28540800015131634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,8,balanced,2.6063626607259116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,8,power_law_1.01,0.18125439882278443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,8,balanced,0.3169440031051636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,8,balanced,0.29340267181396484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,8,power_law_1.01,0.22237439155578614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,8,balanced,0.34382931391398114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,8,power_law_1.01,0.2748863935470581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,8,balanced,0.30780800183614093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,8,power_law_1.01,0.30989439487457277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,8,balanced,0.39587199687957764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,8,power_law_1.01,0.3283967971801758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,8,balanced,0.3450133403142293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,8,power_law_1.01,0.35672318935394287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,8,balanced,0.4466559886932373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,8,power_law_1.01,0.33333759307861327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,8,balanced,0.3784533341725667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,8,power_law_1.01,0.3418623924255371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,8,power_law_1.01,0.3809407949447632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,8,balanced,0.5582026640574137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,8,power_law_1.01,0.35218560695648193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,8,power_law_1.01,0.3634687900543213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,8,balanced,0.47657068570454914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,8,power_law_1.01,0.37813119888305663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,8,power_law_1.01,0.3912895917892456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,8,power_law_1.01,0.3969856023788452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,8,balanced,0.7656853199005127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,8,balanced,0.6769866943359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,8,power_law_1.01,0.42622079849243166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,8,power_law_1.01,0.4981247901916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,8,power_law_1.01,0.505728006362915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,8,power_law_1.01,0.6057727813720704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,8,power_law_1.01,0.689740800857544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,8,balanced,0.9471253554026285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,8,power_law_1.01,0.863315200805664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,8,balanced,0.818458636601766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,8,power_law_1.01,1.0415424346923827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,8,power_law_1.01,1.3971199989318848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,8,power_law_1.01,1.7951040267944336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,8,power_law_1.01,2.063449668884277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,8,power_law_1.01,3.29431037902832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,8,balanced,1.2534879843393962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,8,balanced,4.614730517069499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,8,power_law_1.01,6.217516708374023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,8,balanced,1.0966239770253499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,8,balanced,1.596853256225586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,8,balanced,1.4056320190429688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,8,balanced,2.42303466796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,8,balanced,2.119717280069987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,8,balanced,4.551589330037435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,8,balanced,3.917759895324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,8,power_law_1.01,0.09065600037574768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,8,power_law_1.01,0.12906880378723146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,8,power_law_1.01,0.09937279820442199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,8,power_law_1.01,0.2081216096878052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,8,power_law_1.01,0.13562239408493043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,8,power_law_1.01,0.09248639941215515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,8,power_law_1.01,0.21861119270324708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,8,power_law_1.01,0.09479680061340331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,8,power_law_1.01,0.11925760507583619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,8,power_law_1.01,0.16437120437622071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,8,power_law_1.01,0.16869759559631348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,8,power_law_1.01,0.20173439979553223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,8,power_law_1.01,0.20115199089050292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,8,power_law_1.01,0.20216960906982423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,8,power_law_1.01,0.2197119951248169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,8,power_law_1.01,0.21593599319458007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,8,power_law_1.01,0.21754879951477052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,8,power_law_1.01,0.23002240657806397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,8,power_law_1.01,0.24291839599609374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,8,power_law_1.01,0.2249664068222046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,8,power_law_1.01,0.23924479484558106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,8,power_law_1.01,0.2655679941177368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,8,power_law_1.01,0.29160959720611573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,8,power_law_1.01,0.32565760612487793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,8,power_law_1.01,0.35606400966644286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,8,power_law_1.01,0.450713586807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,8,power_law_1.01,0.5241024017333984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,8,power_law_1.01,0.699782419204712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,8,power_law_1.01,0.8633600234985351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,8,power_law_1.01,1.186150360107422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,8,power_law_1.01,1.4720895767211915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,8,power_law_1.01,1.8633920669555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,8,power_law_1.01,2.8329599380493162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,8,power_law_1.01,5.319660949707031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,8,power_law_1.2,0.12439680099487305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,8,power_law_1.2,0.19637759923934936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,8,power_law_1.2,0.08911359906196595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,8,power_law_1.2,0.16632319688796998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,8,power_law_1.2,0.13066240549087524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,8,power_law_1.2,0.11188479661941528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,8,power_law_1.2,0.12581759691238403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,8,power_law_1.2,0.16246399879455567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,8,power_law_1.2,0.09409919977188111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,8,power_law_1.2,0.20841600894927978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,8,power_law_1.2,0.11367039680480957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,8,power_law_1.2,0.22284159660339356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,8,power_law_1.2,0.13886719942092896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,8,power_law_1.2,0.3044032096862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,8,power_law_1.2,0.16162559986114503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,8,power_law_1.2,0.32255361080169676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,8,power_law_1.2,0.1935487985610962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,8,power_law_1.2,0.31410560607910154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,8,power_law_1.2,0.1953727960586548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,8,power_law_1.2,0.32660479545593263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,8,power_law_1.2,0.20757761001586914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,8,power_law_1.2,0.32522881031036377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,8,power_law_1.2,0.2060863971710205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,8,power_law_1.2,0.3250175952911377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,8,power_law_1.2,0.2183680057525635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,8,power_law_1.2,0.3534656047821045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,8,power_law_1.2,0.22517120838165283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,8,power_law_1.2,0.3764928102493286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,8,power_law_1.2,0.2211519956588745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,8,power_law_1.2,0.37793281078338625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,8,power_law_1.2,0.2403712034225464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,8,power_law_1.2,0.3905855894088745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,8,power_law_1.2,0.24213120937347413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,8,power_law_1.2,0.4214335918426514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,8,power_law_1.2,0.24110720157623292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,8,power_law_1.2,0.4554880142211914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,8,power_law_1.2,0.2841536045074463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,8,power_law_1.2,0.5048960208892822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,8,power_law_1.2,0.28263039588928224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,8,power_law_1.2,0.5524928092956543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,8,power_law_1.2,0.34009599685668945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,8,power_law_1.2,0.647660779953003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,8,power_law_1.2,0.39157121181488036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,8,power_law_1.2,0.7126207828521729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,8,power_law_1.2,0.4853375911712646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,8,power_law_1.2,0.9190272331237793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,8,power_law_1.2,0.5587135791778565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,8,power_law_1.2,1.1627584457397462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,8,power_law_1.2,0.7569536209106446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,8,power_law_1.2,1.5744383811950684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,8,power_law_1.2,0.979468822479248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,8,power_law_1.2,1.9904064178466796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,8,power_law_1.2,1.2753472328186035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,8,power_law_1.2,2.3662336349487303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,8,power_law_1.2,1.6674495697021485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,8,power_law_1.2,1.9879423141479493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,8,power_law_1.2,3.3386302947998048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,8,power_law_1.2,3.0302783966064455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,8,power_law_1.2,6.640767669677734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,8,power_law_1.2,5.956550216674804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,8,power_law_1.01,0.1254207968711853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,8,power_law_1.01,0.15901440382003784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,8,power_law_1.01,0.17229440212249755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,8,power_law_1.01,0.20940160751342773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,8,power_law_1.01,0.21452798843383789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,8,power_law_1.01,0.20208640098571778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,8,power_law_1.01,0.21823999881744385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,8,power_law_1.01,0.22165119647979736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,8,power_law_1.01,0.2200256109237671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,8,power_law_1.01,0.23144960403442383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,8,power_law_1.01,0.23635199069976806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,8,power_law_1.01,0.24469759464263915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,8,power_law_1.01,0.25296640396118164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,8,power_law_1.01,0.26886401176452634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,8,power_law_1.01,0.27492480278015136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,8,power_law_1.2,0.10680960416793824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,8,power_law_1.01,0.31719040870666504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,8,power_law_1.2,0.14691840410232543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,8,power_law_1.01,0.35440640449523925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,8,power_law_1.2,0.15028480291366578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,8,power_law_1.2,0.09948800206184387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,8,power_law_1.01,0.41125121116638186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,8,power_law_1.01,0.4990528106689453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,8,power_law_1.2,0.12495360374450684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,8,power_law_1.2,0.15379199981689454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,8,power_law_1.01,0.6294464111328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,8,power_law_1.2,0.17541760206222534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,8,power_law_1.01,0.7997504234313965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,8,power_law_1.2,0.19997440576553344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,8,power_law_1.01,1.1012991905212401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,8,power_law_1.2,0.20967040061950684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,8,power_law_1.01,1.3758720397949218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,8,power_law_1.2,0.21032960414886476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,8,power_law_1.01,1.675699234008789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,8,power_law_1.2,0.21473920345306396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,8,power_law_1.01,2.452044868469238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,8,power_law_1.2,0.22330238819122314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,8,power_law_1.2,0.22981760501861573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,8,power_law_1.01,5.031475067138672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,8,power_law_1.2,0.23594241142272948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,8,power_law_1.2,0.24158720970153807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,8,power_law_1.2,0.250163197517395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,8,power_law_1.2,0.23821439743041992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,8,power_law_1.2,0.27615358829498293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,8,power_law_1.2,0.2865664005279541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,8,power_law_1.2,0.3329279899597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,8,power_law_1.2,0.37832319736480713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,8,power_law_1.2,0.4636544227600098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,8,power_law_1.2,0.5251967906951904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,8,power_law_1.2,0.6959487915039062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,8,power_law_1.2,0.8332799911499024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,8,power_law_1.2,1.151200008392334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,8,power_law_1.2,1.5494655609130858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,8,power_law_1.2,1.787481689453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,8,power_law_1.2,2.6814271926879885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,8,power_law_1.2,5.706489562988281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,4,balanced,0.05449600021044413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,4,balanced,0.05938666562239329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,4,balanced,0.06465066472689311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,4,balanced,0.07052266597747803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,4,balanced,0.07810666660467784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,4,balanced,0.09219200412432353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,4,balanced,0.09249599774678548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,4,balanced,0.1325973371664683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,4,balanced,0.12731200456619263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,4,balanced,0.2020853360493978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,4,balanced,0.19051732619603476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,4,balanced,0.265994668006897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,4,balanced,0.32519465684890747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,4,balanced,0.25653332471847534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,4,balanced,0.4262186686197917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,4,balanced,0.25276267528533936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,4,balanced,0.41387200355529785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,4,balanced,0.2550453344980876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,4,balanced,0.43693868319193524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,4,balanced,0.2564000089963277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,4,balanced,0.42981334527333576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,4,balanced,0.25895466407140094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,4,balanced,0.42952001094818115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,4,balanced,0.258133331934611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,4,balanced,0.43554667631785077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,4,balanced,0.26714666684468585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,4,balanced,0.41444798310597736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,4,balanced,0.2721173365910848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,4,balanced,0.4225546518961589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,4,balanced,0.2743840018908183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,4,balanced,0.4219413201014201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,4,balanced,0.2773226698239644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,4,balanced,0.430245320002238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,4,balanced,0.2990933259328206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,4,balanced,0.4344480037689209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,4,balanced,0.3126986622810364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,4,balanced,0.45278934637705487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,4,balanced,0.34833065668741864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,4,balanced,0.45721598466237384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,4,balanced,0.3800319830576579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,4,balanced,0.4795573155085246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,4,balanced,0.5070613225301107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,4,balanced,0.4516586860020955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,4,balanced,0.5558186769485474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,4,balanced,0.5193119843800863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,4,balanced,0.6075573364893595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,4,balanced,0.6625973383585612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,4,balanced,0.7183573246002197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,4,balanced,0.8856639862060547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,4,balanced,0.9796586831410726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,4,balanced,1.1446932951609294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,4,balanced,1.149888038635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,4,balanced,0.0645066648721695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,4,balanced,0.07214400172233582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,4,balanced,0.08358400066693623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,4,balanced,0.10830932855606079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,4,balanced,1.5062559445699055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,4,balanced,0.15441067020098367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,4,balanced,0.236026664574941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,4,balanced,0.3116640051205953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,4,balanced,0.2969759901364644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,4,balanced,0.2908160090446472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,4,balanced,0.28651734193166095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,4,balanced,0.286682665348053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,4,balanced,1.5212000211079915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,4,balanced,0.2802453239758809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,4,balanced,0.28334399064381915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,4,balanced,0.28431467215220135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,4,balanced,0.29736532767613727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,4,balanced,0.2938506603240967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,4,balanced,1.88702392578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,4,balanced,0.304202675819397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,4,balanced,0.3091946641604106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,4,balanced,0.3223573366800944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,4,balanced,0.3505653142929077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,4,balanced,0.37347201506296795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,4,power_law_1.01,0.09642239809036254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,4,balanced,1.971552054087321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,4,power_law_1.01,0.12225919961929321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,4,power_law_1.01,0.18486399650573732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,4,balanced,0.42607998847961426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,4,power_law_1.01,0.10458879470825196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,4,power_law_1.01,0.16851840019226075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,4,power_law_1.01,0.20588159561157227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,4,balanced,0.4737226565678914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,4,power_law_1.01,0.24763519763946534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,4,power_law_1.01,0.29859840869903564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,4,power_law_1.01,0.29843199253082275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,4,balanced,2.88150946299235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,4,balanced,0.5952106714248657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,4,power_law_1.01,0.31607680320739745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,4,power_law_1.01,0.32305920124053955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,4,power_law_1.01,0.350272011756897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,4,power_law_1.01,0.33133440017700194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,4,power_law_1.01,0.3530944108963013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,4,balanced,0.857754627863566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,4,power_law_1.01,0.37944319248199465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,4,power_law_1.01,0.3624576091766357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,4,power_law_1.01,0.38623359203338625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,4,balanced,2.923583984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,4,power_law_1.01,0.38857600688934324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,4,power_law_1.01,0.42539520263671876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,4,power_law_1.01,0.46492800712585447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,4,balanced,1.0502613385518391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,4,power_law_1.01,0.5178624153137207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,4,power_law_1.01,0.6056575775146484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,4,power_law_1.01,0.7038656234741211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,4,power_law_1.01,0.9393535614013672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,4,power_law_1.01,1.0712703704833983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,4,power_law_1.01,1.4796799659729003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,4,power_law_1.01,1.814476776123047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,4,power_law_1.01,2.148908805847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,4,balanced,1.4143999417622883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,4,power_law_1.01,3.2506431579589843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,4,power_law_1.01,6.5507965087890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,4,balanced,5.49562136332194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,4,balanced,1.8050079345703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,4,balanced,5.42197863260905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,4,balanced,2.7098026275634766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,4,balanced,5.12938117980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,4,power_law_1.01,0.07601919770240784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,4,power_law_1.01,0.09493119716644287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,4,power_law_1.01,0.1310912013053894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,4,power_law_1.01,0.08603519797325135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,4,power_law_1.01,0.08370559811592101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,4,power_law_1.01,0.11408640146255493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,4,power_law_1.01,0.12039680480957031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,4,power_law_1.01,0.15843199491500853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,4,power_law_1.01,0.14729599952697753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,4,power_law_1.01,0.10437120199203491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,4,power_law_1.01,0.16392960548400878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,4,power_law_1.01,0.13111040592193604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,4,power_law_1.01,0.19259519577026368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,4,power_law_1.01,0.16220799684524537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,4,power_law_1.01,0.20384640693664552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,4,power_law_1.01,0.17604479789733887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,4,power_law_1.01,0.20449280738830566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,4,power_law_1.01,0.20864639282226563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,4,power_law_1.01,0.21319680213928222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,4,power_law_1.01,0.22464640140533448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,4,power_law_1.01,0.23585278987884523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,4,power_law_1.2,0.09796479940414429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,4,power_law_1.01,0.24391040802001954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,4,power_law_1.2,0.12392959594726563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,4,power_law_1.01,0.2479680061340332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,4,power_law_1.2,0.11980160474777221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,4,power_law_1.01,0.25639679431915285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,4,power_law_1.2,0.1100160002708435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,4,power_law_1.2,0.16753920316696166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,4,power_law_1.01,0.27482240200042723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,4,power_law_1.2,0.20627200603485107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,4,power_law_1.01,0.29459199905395506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,4,power_law_1.2,0.22917120456695556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,4,power_law_1.01,0.34311039447784425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,4,power_law_1.2,0.2932607889175415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,4,power_law_1.01,0.3829184055328369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,4,power_law_1.2,0.2966399908065796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,4,power_law_1.01,0.4819392204284668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,4,power_law_1.2,0.3154623985290527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,4,power_law_1.01,0.5874559879302979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,4,power_law_1.2,0.31683199405670165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,4,power_law_1.01,0.7724031925201416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,4,power_law_1.2,0.33549439907073975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,4,power_law_1.01,0.9707263946533203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,4,power_law_1.2,0.34069759845733644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,4,power_law_1.01,1.3134719848632812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,4,power_law_1.2,0.35593600273132325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,4,power_law_1.01,1.7192064285278321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,4,power_law_1.2,0.3860608100891113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,4,power_law_1.01,2.0346303939819337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,4,power_law_1.2,0.3797055959701538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,4,power_law_1.01,3.217497634887695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,4,power_law_1.2,0.390995192527771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,4,power_law_1.2,0.4048768043518066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,4,power_law_1.01,6.187340927124024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,4,power_law_1.2,0.44237442016601564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,4,power_law_1.2,0.5003583908081055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,4,power_law_1.2,0.5469696044921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,4,power_law_1.2,0.6594880104064942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,4,power_law_1.2,0.7352128028869629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,4,power_law_1.2,0.948031997680664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,4,power_law_1.2,1.1143808364868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,4,power_law_1.2,1.5613759994506835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,4,power_law_1.2,1.9029760360717773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,4,power_law_1.2,2.2069631576538087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,4,power_law_1.2,3.555750274658203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,4,power_law_1.2,7.193401336669922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,4,power_law_1.2,0.0726207971572876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,4,power_law_1.2,0.0946943998336792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,4,power_law_1.2,0.08727040290832519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,4,power_law_1.2,0.08595200181007386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,4,power_law_1.2,0.11616640090942383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,4,power_law_1.2,0.13214720487594606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,4,power_law_1.2,0.1594879984855652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,4,power_law_1.2,0.18486399650573732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,4,power_law_1.2,0.19228800535202026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,4,power_law_1.2,0.1918015956878662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,4,power_law_1.2,0.2093503952026367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,4,power_law_1.2,0.2133568048477173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,4,power_law_1.2,0.21514880657196045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,4,power_law_1.2,0.22338559627532958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,4,power_law_1.2,0.23800320625305177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,4,power_law_1.2,0.23315839767456054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,4,power_law_1.2,0.26089599132537844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,4,power_law_1.2,0.2823359966278076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,4,power_law_1.2,0.30773119926452636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,4,power_law_1.2,0.3571968078613281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,4,power_law_1.2,0.4157120227813721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,4,power_law_1.2,0.49923200607299806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,4,power_law_1.2,0.6150271892547607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,4,power_law_1.2,0.7939328193664551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,4,power_law_1.2,0.9745471954345704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,4,power_law_1.2,1.3635264396667481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,4,power_law_1.2,1.6501760482788086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,4,power_law_1.2,2.1968255996704102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,4,power_law_1.2,3.372275161743164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,4,power_law_1.2,6.189926528930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,4,power_law_1.01,0.22420480251312255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,4,power_law_1.01,0.23916800022125245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,4,power_law_1.01,0.24043519496917726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,4,power_law_1.01,0.23400321006774902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,4,power_law_1.01,0.24303359985351564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,4,power_law_1.01,0.24775679111480714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,4,power_law_1.01,0.2493760108947754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,4,power_law_1.01,0.26598401069641114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,4,power_law_1.2,0.08485760092735291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,4,power_law_1.01,0.2670783996582031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,4,power_law_1.2,0.11272319555282592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,4,power_law_1.01,0.2708735942840576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,4,power_law_1.2,0.09909120202064514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,4,power_law_1.01,0.2879744052886963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,4,power_law_1.01,0.30986239910125735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,4,power_law_1.2,0.10012160539627075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,4,power_law_1.01,0.3557503938674927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,4,power_law_1.2,0.1263808012008667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,4,power_law_1.01,0.3982975959777832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,4,power_law_1.2,0.15686399936676027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,4,power_law_1.2,0.17582080364227295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,4,power_law_1.01,0.46817917823791505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,4,power_law_1.2,0.2100543975830078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,4,power_law_1.01,0.5708608150482177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,4,power_law_1.2,0.21418240070343017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,4,power_law_1.01,0.7146048069000244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,4,power_law_1.2,0.22588160037994384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,4,power_law_1.01,0.8969152450561524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,4,power_law_1.2,0.23634560108184816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,4,power_law_1.01,1.244211196899414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,4,power_law_1.2,0.24341120719909667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,4,power_law_1.01,1.602137565612793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,4,power_law_1.2,0.243174409866333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,4,power_law_1.01,1.970195198059082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,4,power_law_1.2,0.24840960502624512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,4,power_law_1.01,3.2577793121337892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,2,balanced,0.07100800176461537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,4,power_law_1.2,0.25501439571380613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,2,balanced,0.08306133250395457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,2,balanced,0.09847467144330342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,4,power_law_1.01,5.5583038330078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,4,power_law_1.2,0.26537599563598635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,2,balanced,0.12847466270128885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,2,balanced,0.20003734032313028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,4,power_law_1.2,0.2862400054931641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,2,balanced,0.32256533702214557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,4,power_law_1.2,0.30475521087646484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,2,balanced,0.4580693244934082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,2,balanced,0.4463413159052531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,4,power_law_1.2,0.3290623903274536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,2,balanced,0.45448533693949383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,4,power_law_1.2,0.3716415882110596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,2,balanced,0.4533066749572754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,2,balanced,0.44974398612976074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,4,power_law_1.2,0.4157120227813721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,2,balanced,0.45123199621836346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,4,power_law_1.2,0.5135744094848633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,2,balanced,0.44790931542714435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,4,power_law_1.2,0.5973696231842041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,2,balanced,0.44926400979359943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,2,balanced,0.45573333899180096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,4,power_law_1.2,0.7875840187072753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,2,balanced,0.45338133970896405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,4,power_law_1.2,0.990227222442627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,2,balanced,0.47016000747680664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,4,power_law_1.2,1.3564864158630372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,2,balanced,0.4913333257039388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,4,power_law_1.2,1.6689535140991212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,2,balanced,0.5056586662928263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,4,power_law_1.2,2.1176191329956056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,4,power_law_1.2,3.34851188659668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,2,balanced,0.5522666772206625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,4,power_law_1.2,6.196672058105468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,2,balanced,0.5853600104649862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,2,balanced,0.6703840096791586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,2,balanced,0.7497813701629639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,2,balanced,0.9190773169199625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,2,balanced,1.268719991048177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,2,balanced,0.06281066437562306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,2,balanced,0.067930668592453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,2,balanced,0.08004799981911977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,2,balanced,0.10728533069292705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,2,balanced,0.15386666854222616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,2,balanced,0.2323253353436788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,2,balanced,0.33213333288828534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,2,balanced,0.3097440004348755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,2,balanced,1.594314734141032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,2,balanced,0.3102239966392517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,2,balanced,0.3133653402328491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,2,balanced,0.31774399677912396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,2,balanced,0.3124000032742818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,2,balanced,0.3145493268966675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,2,balanced,0.31331733862559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,2,balanced,0.32871466875076294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,2,balanced,0.3259999950726827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,2,balanced,0.3423573176066081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,2,balanced,0.3662826617558797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,2,balanced,0.3917333285013835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,2,balanced,0.44252268473307294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,2,balanced,2.0817012786865234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,2,balanced,0.48286934693654376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,2,balanced,0.5830239852269491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,2,balanced,0.6741546789805094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,2,balanced,0.8804852962493896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,2,balanced,0.07895466685295105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,2,balanced,2.60588804880778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,2,balanced,0.08400000135103862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,2,balanced,0.09801066915194194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,2,balanced,0.13594133655230203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,2,balanced,0.19223467508951822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,2,balanced,0.28887999057769775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,2,balanced,1.1957653363545735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,2,balanced,0.3736533323923747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,2,balanced,0.37789865334828693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,2,balanced,0.3633013168970744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,2,balanced,0.36829864978790283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,2,balanced,0.37460800011952716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,2,balanced,0.3709706862767537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,2,balanced,0.36847468217213947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,2,balanced,0.37201066811879474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,2,balanced,0.3924586772918701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,2,balanced,1.5855733553568523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,2,balanced,0.38625065485636395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,2,balanced,0.40824000040690106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,2,balanced,0.4188586473464966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,2,balanced,0.43435200055440265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,2,balanced,0.4799199899037679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,2,balanced,0.5179520050684611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,2,balanced,3.9023307164510093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,2,balanced,2.0818400382995605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,2,balanced,0.5960586468378702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,2,power_law_1.01,0.08181759715080261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,2,power_law_1.01,0.10094720125198364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,2,balanced,0.6831626892089844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,2,power_law_1.01,0.13030400276184081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,2,power_law_1.01,0.11380480527877808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,2,power_law_1.01,0.1763327956199646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,2,power_law_1.01,0.21502718925476075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,2,balanced,0.8631306489308676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,2,power_law_1.01,0.2488255977630615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,2,power_law_1.01,0.3072767972946167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,2,power_law_1.01,0.3201920032501221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,2,power_law_1.01,0.32639360427856445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,2,balanced,2.5887680053710938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,2,balanced,1.237663984298706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,2,power_law_1.01,0.33793280124664304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,2,power_law_1.01,0.35043840408325194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,2,power_law_1.01,0.35540480613708497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,2,power_law_1.01,0.3689919948577881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,2,power_law_1.01,0.3848383903503418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,2,power_law_1.01,0.3826688051223755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,2,power_law_1.01,0.405625581741333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,2,balanced,1.5678186416625977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,2,power_law_1.01,0.4376383781433105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,2,power_law_1.01,0.46929278373718264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,2,power_law_1.01,0.5085567951202392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,2,power_law_1.01,0.5855231761932373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,2,power_law_1.01,0.6973567962646484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,2,power_law_1.01,0.822662353515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,2,power_law_1.01,1.0302847862243651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,2,balanced,2.099621295928955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,2,power_law_1.01,1.2715456008911132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,2,power_law_1.01,1.6985279083251954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,2,balanced,3.9737065633138022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,2,power_law_1.01,2.1822015762329103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,2,power_law_1.01,2.5730815887451173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,2,power_law_1.01,4.098406219482422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,2,power_law_1.01,7.966099548339844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,2,balanced,7.263194402058919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,2,balanced,2.6262079874674478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,2,balanced,4.025055885314941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,2,balanced,7.609200159708659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,2,balanced,7.721466700236003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,2,power_law_1.01,0.06752640008926392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,2,power_law_1.01,0.08629760146141052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,2,power_law_1.01,0.08357759714126586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,2,power_law_1.01,0.10030080080032348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,2,power_law_1.01,0.10460799932479858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,2,power_law_1.01,0.13122559785842897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,2,power_law_1.01,0.09718400239944458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,2,power_law_1.01,0.1177791953086853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,2,power_law_1.01,0.13639039993286134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,2,power_law_1.01,0.15837440490722657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,2,power_law_1.01,0.1601088047027588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,2,power_law_1.01,0.20472960472106932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,2,power_law_1.01,0.18456319570541382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,2,power_law_1.01,0.21962239742279052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,2,power_law_1.01,0.2221440076828003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,2,power_law_1.01,0.22740480899810792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,2,power_law_1.01,0.27080960273742677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,2,power_law_1.01,0.23624958992004394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,2,power_law_1.01,0.2750591993331909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,2,power_law_1.01,0.2409600019454956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,2,power_law_1.01,0.29402880668640136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,2,power_law_1.01,0.2529344081878662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,2,power_law_1.01,0.2974656105041504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,2,power_law_1.01,0.26042881011962893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,2,power_law_1.01,0.3058880090713501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,2,power_law_1.01,0.266374397277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,2,power_law_1.01,0.3083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,2,power_law_1.01,0.28523519039154055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,2,power_law_1.01,0.3220223903656006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,2,power_law_1.01,0.28511359691619875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,2,power_law_1.01,0.3286272048950195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,2,power_law_1.01,0.3034303903579712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,2,power_law_1.01,0.33799679279327394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,2,power_law_1.01,0.3297663927078247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,2,power_law_1.01,0.3513792037963867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,2,power_law_1.01,0.3605439901351929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,2,power_law_1.01,0.3709439992904663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,2,power_law_1.01,0.42327041625976564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,2,power_law_1.01,0.4086336135864258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,2,power_law_1.01,0.4842688083648682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,2,power_law_1.01,0.4704319953918457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,2,power_law_1.01,0.6028223991394043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,2,power_law_1.01,0.5358016014099121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,2,power_law_1.01,0.7258687973022461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,2,power_law_1.01,0.6487040042877197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,2,power_law_1.01,0.9703616142272949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,2,power_law_1.01,0.7651968002319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,2,power_law_1.01,1.2405055999755858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,2,power_law_1.01,1.0225279808044434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,2,power_law_1.01,1.2556927680969239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,2,power_law_1.01,1.7327999114990233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,2,power_law_1.01,1.739187240600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,2,power_law_1.01,2.1461183547973635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,2,power_law_1.01,2.1859392166137694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,2,power_law_1.01,2.678175926208496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,2,power_law_1.2,0.07983999848365783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,2,power_law_1.01,2.7668928146362304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,2,power_law_1.2,0.10083199739456176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,2,power_law_1.01,4.049785614013672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,2,power_law_1.2,0.09423360228538513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,2,power_law_1.01,4.059699249267578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,2,power_law_1.01,8.103968048095703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,2,power_law_1.2,0.11512960195541382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,2,power_law_1.01,8.108895874023437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,2,power_law_1.2,0.16278400421142578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,2,power_law_1.2,0.18765439987182617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,2,power_law_1.2,0.23820159435272217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,2,power_law_1.2,0.29386239051818847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,2,power_law_1.2,0.3097088098526001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,2,power_law_1.2,0.3180351972579956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,2,power_law_1.2,0.327129602432251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,2,power_law_1.2,0.3389951944351196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,2,power_law_1.2,0.34897279739379883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,2,power_law_1.2,0.3595968008041382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,2,power_law_1.2,0.371891188621521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,2,power_law_1.2,0.0689087986946106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,2,power_law_1.2,0.3821248054504395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,2,power_law_1.2,0.08261759877204895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,2,power_law_1.2,0.41074562072753906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,2,power_law_1.2,0.0803264021873474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,2,power_law_1.2,0.4519167900085449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,2,power_law_1.2,0.09807360172271729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,2,power_law_1.2,0.4850624084472656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,2,power_law_1.2,0.12311040163040161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,2,power_law_1.2,0.5399040222167969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,2,power_law_1.2,0.14975359439849853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,2,power_law_1.2,0.5982783794403076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,2,power_law_1.2,0.7177919864654541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,2,power_law_1.2,0.16853760480880736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,2,power_law_1.2,0.2023871898651123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,2,power_law_1.2,0.8370112419128418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,2,power_law_1.2,0.22310400009155273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,2,power_law_1.2,1.1035264015197754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,2,power_law_1.2,0.22918400764465333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,2,power_law_1.2,1.3418047904968262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,2,power_law_1.2,0.2342400074005127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,2,power_law_1.2,1.7716480255126954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,2,power_law_1.2,0.2502912044525146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,2,power_law_1.2,2.2953983306884767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,2,power_law_1.2,0.2543168067932129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,2,power_law_1.2,2.7545024871826174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,2,power_law_1.2,0.2641727924346924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,2,power_law_1.2,4.1561279296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,2,power_law_1.2,0.2662719964981079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,2,power_law_1.2,0.28935039043426514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,2,power_law_1.2,7.9542594909667965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,2,power_law_1.2,0.3023808002471924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,2,power_law_1.2,0.33323519229888915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,2,power_law_1.2,0.3646656036376953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,2,power_law_1.2,0.4348480224609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,2,power_law_1.2,0.5009535789489746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,2,power_law_1.2,0.6195199966430665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,2,power_law_1.2,0.7495039939880371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,2,power_law_1.2,1.0077504158020019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,2,power_law_1.2,1.2365568161010743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,2,power_law_1.2,1.7503423690795898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,2,power_law_1.2,2.1673791885375975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,2,power_law_1.2,2.719264030456543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,2,power_law_1.2,4.120806503295898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,2,power_law_1.2,7.877708435058594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,2,power_law_1.2,0.0870464026927948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,2,power_law_1.2,0.09767040014266967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,2,power_law_1.2,0.10830719470977783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,2,power_law_1.2,0.11754879951477051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,1,balanced,0.08191466828187306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,2,power_law_1.2,0.15016959905624389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,1,balanced,0.09601599971453349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,2,power_law_1.2,0.1865023970603943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,1,balanced,0.12061867117881775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,1,balanced,0.16048533717791238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,2,power_law_1.2,0.21021440029144287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,1,balanced,0.23745065927505493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,2,power_law_1.2,0.25817599296569826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,1,balanced,0.3985919952392578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,1,balanced,0.5574026505152384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,2,power_law_1.2,0.2788160085678101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,1,balanced,0.5468480189641317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,2,power_law_1.2,0.27783679962158203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,1,balanced,0.549018661181132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,1,balanced,0.539242664972941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,2,power_law_1.2,0.28555519580841066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,1,balanced,0.5476053158442179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,2,power_law_1.2,0.30633599758148194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,1,balanced,0.554090658823649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,2,power_law_1.2,0.3084800004959106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,1,balanced,0.5529599984486898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,2,power_law_1.2,0.31303040981292723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,1,balanced,0.5565013488133749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,1,balanced,0.5671413342157999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,2,power_law_1.2,0.3253567934036255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,1,balanced,0.5680319865544637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,2,power_law_1.2,0.33773438930511473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,1,balanced,0.5860746701558431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,2,power_law_1.2,0.35386879444122316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,2,power_law_1.2,0.38071680068969727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,1,balanced,0.6315199931462606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,2,power_law_1.2,0.427564811706543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,1,balanced,0.6617813507715861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,2,power_law_1.2,0.49276161193847656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,2,power_law_1.2,0.5490943908691406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,1,balanced,0.7306079864501953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,2,power_law_1.2,0.676691198348999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,1,balanced,0.7927253246307373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,2,power_law_1.2,0.8110400199890136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,2,power_law_1.2,1.0530240058898925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,2,power_law_1.2,1.3315263748168946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,1,balanced,0.929370641708374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,2,power_law_1.2,1.8506879806518555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,2,power_law_1.2,2.2566911697387697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,2,power_law_1.2,2.7554048538208007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,1,balanced,1.0577173233032227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,2,power_law_1.2,4.32922248840332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,1,balanced,0.07525333265463512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,2,power_law_1.2,8.427865600585937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,1,balanced,0.07965333263079326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,1,balanced,0.09603200356165568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,1,balanced,1.374735991160075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,1,balanced,0.13752532998720804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,1,balanced,0.19112000862757364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,1,balanced,0.2983146707216899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,1,balanced,0.40595734119415283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,1,balanced,0.4068853457768758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,1,balanced,0.4102880160013835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,1,balanced,0.40990400314331055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,1,balanced,0.4204853375752767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,1,balanced,1.8530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,1,balanced,0.4260053237279256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,1,balanced,0.4241120020548503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,1,balanced,0.4352960189183553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,1,balanced,0.4432213306427002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,1,balanced,0.44654401143391925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,1,balanced,0.4603840112686157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,1,balanced,0.5128320058186849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,1,balanced,0.11521066228548686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,1,balanced,0.5437386830647787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,1,balanced,0.12495467066764832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,1,balanced,0.15222400426864624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,1,balanced,2.4249226252237954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,1,balanced,0.1998400092124939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,1,balanced,0.6258560021718343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,1,balanced,0.2884213328361511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,1,balanced,0.4442559878031413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,1,balanced,0.618938684463501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,1,balanced,0.6897226969401041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,1,balanced,0.6091573238372803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,1,balanced,0.6031146844228109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,1,balanced,0.6016586621602377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,1,balanced,0.839786688486735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,1,balanced,0.5954986810684204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,1,balanced,0.5962560176849365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,1,balanced,0.5986506541570028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,1,balanced,0.6105120182037354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,1,balanced,0.9832586447397867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,1,balanced,0.625711997350057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,1,balanced,0.6209226846694946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,1,balanced,0.6452266772588094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,1,balanced,3.2126506169637046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,1,balanced,0.6833493709564209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,1,balanced,1.3060373465220134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,1,balanced,0.7151626745859782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,1,balanced,0.80076797803243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,1,balanced,0.8718079725901285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,1,balanced,1.7698826789855957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,1,balanced,1.0331253210703533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,1,balanced,1.1954399744669597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,1,balanced,3.9454345703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,1,balanced,2.373477300008138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,1,balanced,1.5218399365743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,1,balanced,2.1518239974975586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,1,balanced,3.1738719940185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,1,balanced,2.807861328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,1,balanced,6.000037511189778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,1,balanced,3.9410667419433594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,1,balanced,3.7455838521321616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,1,balanced,4.672042528788249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,1,power_law_1.01,0.07545599937438965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,1,balanced,6.061029434204102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,1,power_law_1.01,0.0871936023235321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,1,power_law_1.01,0.11290240287780762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,1,power_law_1.01,0.13300479650497438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,1,power_law_1.01,0.20135679244995117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,1,power_law_1.01,0.24367361068725585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,1,power_law_1.01,0.28958079814910886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,1,power_law_1.01,0.3801408052444458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,1,power_law_1.01,0.3808896064758301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,1,power_law_1.01,0.3871999979019165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,1,balanced,7.170762379964192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,1,power_law_1.01,0.39835519790649415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,1,power_law_1.01,0.4160575866699219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,1,power_law_1.01,0.43030400276184083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,1,power_law_1.01,0.4447616100311279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,1,power_law_1.01,0.46329598426818847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,1,balanced,11.260106404622396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,1,power_law_1.01,0.4592576026916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,1,power_law_1.01,0.4920191764831543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,1,power_law_1.01,0.5570879936218261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,1,power_law_1.01,0.5927743911743164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,1,power_law_1.01,0.6841536045074463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,1,power_law_1.01,0.7613183975219726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,1,power_law_1.01,0.9324864387512207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,1,power_law_1.01,1.1007360458374023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,1,power_law_1.01,1.4329536437988282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,1,power_law_1.01,1.7721920013427734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,1,power_law_1.01,2.4317440032958983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,1,power_law_1.01,3.093471908569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,1,power_law_1.01,3.7804927825927734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,1,power_law_1.01,5.791110229492188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,1,balanced,11.697226206461588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,1,power_law_1.01,11.203392028808594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,1,balanced,13.689109802246094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,1,power_law_1.01,0.06885759830474854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,1,power_law_1.01,0.07573760151863099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,1,power_law_1.01,0.08813440203666686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,1,power_law_1.01,0.1124351978302002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,1,power_law_1.01,0.15687040090560914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,1,power_law_1.01,0.2025984048843384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,1,power_law_1.01,0.23123838901519775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,1,power_law_1.01,0.28450560569763184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,1,power_law_1.01,0.29275519847869874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,1,power_law_1.01,0.30243840217590334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,1,power_law_1.01,0.31405439376831057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,1,power_law_1.01,0.33048961162567136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,1,power_law_1.01,0.3345151901245117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,1,power_law_1.01,0.3573888063430786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,1,power_law_1.01,0.3655231952667236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,1,power_law_1.01,0.3804352045059204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,1,power_law_1.01,0.40380158424377444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,1,power_law_1.01,0.45265917778015136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,1,power_law_1.01,0.49364480972290037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,1,power_law_1.01,0.10945279598236084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,1,power_law_1.01,0.5898431777954102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,1,power_law_1.2,0.07656319737434387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,1,power_law_1.01,0.12216320037841796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,1,power_law_1.01,0.6736256122589112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,1,power_law_1.2,0.08902400135993957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,1,power_law_1.01,0.8541440010070801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,1,power_law_1.01,0.1487104058265686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,1,power_law_1.2,0.10398720502853394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,1,power_law_1.01,1.0284799575805663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,1,power_law_1.01,0.17736320495605468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,1,power_law_1.2,0.12960000038146974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,1,power_law_1.01,0.2504319906234741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,1,power_law_1.01,1.3723391532897948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,1,power_law_1.2,0.18320000171661377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,1,power_law_1.2,0.06866559982299805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,1,power_law_1.01,0.3103487968444824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,1,power_law_1.01,1.7238975524902345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,1,power_law_1.2,0.2307136058807373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,1,power_law_1.2,0.0772607982158661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,1,power_law_1.01,0.34295680522918703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,1,power_law_1.01,2.4228799819946287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,1,power_law_1.2,0.27422080039978025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,1,power_law_1.2,0.08613759875297547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,1,power_law_1.01,0.41025919914245607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,1,power_law_1.2,0.3503040075302124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,1,power_law_1.01,3.123200035095215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,1,power_law_1.2,0.11382399797439575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,1,power_law_1.01,0.4281919956207275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,1,power_law_1.2,0.36814079284667967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,1,power_law_1.01,3.8402240753173826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,1,power_law_1.2,0.14715520143508912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,1,power_law_1.01,0.43709440231323243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,1,power_law_1.2,0.37239038944244385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,1,power_law_1.2,0.1807487964630127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,1,power_law_1.01,5.952064132690429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,1,power_law_1.01,0.46316800117492674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,1,power_law_1.2,0.39290239810943606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,1,power_law_1.2,0.20304639339447023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,1,power_law_1.01,0.4752768039703369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,1,power_law_1.2,0.4103424072265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,1,power_law_1.01,11.609394836425782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,1,power_law_1.2,0.2723392009735107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,1,power_law_1.01,0.4841728210449219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,1,power_law_1.2,0.41617918014526367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,1,power_law_1.01,0.5089983940124512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,1,power_law_1.2,0.28440959453582765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,1,power_law_1.2,0.4375936031341553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,1,power_law_1.01,0.5205376148223877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,1,power_law_1.2,0.29938559532165526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,1,power_law_1.2,0.4515071868896484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,1,power_law_1.01,0.5352831840515136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,1,power_law_1.2,0.308620810508728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,1,power_law_1.2,0.4667776107788086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,1,power_law_1.01,0.5499328136444092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,1,power_law_1.2,0.3245311975479126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,1,power_law_1.2,0.501087999343872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,1,power_law_1.01,0.6118527889251709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,1,power_law_1.2,0.5739391803741455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,1,power_law_1.2,0.3304383993148804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.01,0.6526080131530761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,1,power_law_1.2,0.5937983989715576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,1,power_law_1.2,0.3499648094177246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.01,0.7652736186981202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,1,power_law_1.2,0.712332820892334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,1,power_law_1.2,0.36362879276275634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.01,0.8647680282592773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,1,power_law_1.2,0.7798719882965088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,1,power_law_1.2,0.3775808095932007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.01,1.0651391983032226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,1,power_law_1.2,0.9557375907897949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,1,power_law_1.2,0.40499200820922854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.01,1.2775551795959472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,1,power_law_1.2,1.1228927612304687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,1,power_law_1.2,0.462553596496582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.01,1.6884416580200194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,1,power_law_1.2,1.459615993499756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,1,power_law_1.2,0.5030079841613769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,1,power_law_1.01,2.1129152297973635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,1,power_law_1.2,1.7879552841186523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,1,power_law_1.2,0.612825584411621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,1,power_law_1.01,2.94073600769043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,1,power_law_1.2,2.4543359756469725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,1,power_law_1.2,0.6971392154693603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,1,power_law_1.01,3.7666240692138673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,1,power_law_1.2,3.1299327850341796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,1,power_law_1.2,0.874880027770996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,1,power_law_1.01,4.593798446655273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,1,power_law_1.2,3.7985855102539063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,1,power_law_1.2,1.0487615585327148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,1,power_law_1.01,7.044371032714844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,1,power_law_1.2,5.8157695770263675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,1,power_law_1.2,1.4034432411193847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,1,power_law_1.2,1.7437311172485352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,1,power_law_1.2,11.203743743896485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,1,power_law_1.01,13.632754516601562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,1,power_law_1.2,2.4381120681762694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,1,power_law_1.2,3.147577667236328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,1,power_law_1.2,3.855487823486328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,1,power_law_1.2,5.954163360595703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,1,power_law_1.2,11.601900482177735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,8,balanced,0.05034666756788889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,1,power_law_1.2,0.11087360382080078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,8,balanced,0.07108800113201141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,8,balanced,0.05400000015894572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,8,balanced,0.07099199791749318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,1,power_law_1.2,0.12400000095367432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,8,balanced,0.04914666712284088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,8,balanced,0.06807999809583028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,8,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,1,power_law_1.2,0.13633919954299928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,8,balanced,0.07293333113193512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,8,balanced,0.05355200171470642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,8,balanced,0.07267733414967854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,1,power_law_1.2,0.1747007966041565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,8,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,8,balanced,0.06963199873765309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,8,balanced,0.05172266562779745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,1,power_law_1.2,0.22933759689331054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,8,balanced,0.06879466772079468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,8,balanced,0.05142400165398916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,8,balanced,0.06736533343791962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,8,balanced,0.05235200126965841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,1,power_law_1.2,0.27587840557098386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,8,balanced,0.06944533189137776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,8,balanced,0.051541333397229515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,1,power_law_1.2,0.30531198978424073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,8,balanced,0.0662773350874583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,8,balanced,0.051167999704678856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,8,balanced,0.04905599852403005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,8,balanced,0.06637866795063019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,1,power_law_1.2,0.41928958892822266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,8,balanced,0.05014933149019877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,8,balanced,0.06573333342870076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,1,power_law_1.2,0.4154047966003418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,8,balanced,0.05128000179926554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,8,balanced,0.061797335743904114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,8,balanced,0.049045334259668984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,8,balanced,0.060778667529424034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,1,power_law_1.2,0.43485441207885744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,8,balanced,0.05086400111516317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,8,balanced,0.0614026685555776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,1,power_law_1.2,0.4600192070007324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,8,balanced,0.05251200000445048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,8,balanced,0.05958400170008341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,8,balanced,0.048207998275756836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,8,balanced,0.05931200087070465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,1,power_law_1.2,0.4658048152923584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,8,balanced,0.06080533564090729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,8,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,8,balanced,0.05709333221117655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,8,balanced,0.04915733138720194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,8,balanced,0.07222400108973186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,1,power_law_1.2,0.47867522239685056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,8,balanced,0.04924266537030538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,8,balanced,0.07701866825421651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,8,balanced,0.07201066613197327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,1,power_law_1.2,0.5140416145324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,8,balanced,0.04952000081539154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,8,balanced,0.08360000451405843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,8,balanced,0.08648533622423808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,8,balanced,0.04916266600290934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,1,power_law_1.2,0.5148416042327881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,8,balanced,0.05128000179926554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,8,balanced,0.08981866637865703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,8,balanced,0.117658664782842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,8,balanced,0.05109866460164388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,1,power_law_1.2,0.5276095867156982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,8,balanced,0.04960533479849497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,8,balanced,0.11860799789428711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,8,balanced,0.13290666540463766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,1,power_law_1.2,0.5566592216491699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,8,balanced,0.04888000090916952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,8,balanced,0.048751999934514366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,8,balanced,0.12973333398501077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,1,power_law_1.2,0.6165184020996094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,8,balanced,0.1823520064353943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,8,balanced,0.051488002141316734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,8,balanced,0.04881600042184194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.2,0.6830592155456543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,8,balanced,0.17621866861979166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,8,balanced,0.04900266726811727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,8,balanced,0.22116265694300333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.2,0.7970880031585693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,8,balanced,0.05073066552480062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,8,balanced,0.04874666531880697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.2,0.8997376441955567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,8,balanced,0.2135093410809835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,8,balanced,0.049039999643961586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,8,balanced,0.30984000364939374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.2,1.1004032135009765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,8,balanced,0.06087466577688853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,8,balanced,0.058746665716171265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.2,1.317363166809082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,8,balanced,0.3022720019022624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,8,balanced,0.06703466673692067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.2,1.732588768005371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,8,balanced,0.3961919943491618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,8,balanced,0.07354133327802022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,1,power_law_1.2,2.146784019470215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,8,balanced,0.09158933162689209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,1,power_law_1.2,2.9653120040893555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,8,balanced,0.3762133518854777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,8,balanced,0.11265599727630615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,8,balanced,0.4806026617685954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,1,power_law_1.2,3.79290885925293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,8,balanced,0.14307199915250143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,1,power_law_1.2,4.607084655761719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,8,balanced,0.4642346700032552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,1,power_law_1.2,7.0631553649902346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,8,balanced,0.17091200749079385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,8,balanced,0.7417600154876709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,1,power_law_1.2,13.632762145996093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,8,balanced,0.23774933815002441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,8,balanced,0.7090453306833903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,8,balanced,0.30152533451716107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,8,balanced,1.4367733001708984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,8,balanced,0.3607786496480306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,8,balanced,1.3622612953186035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,8,balanced,0.5465759833653768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,8,balanced,1.0467627048492432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,8,balanced,0.0215786670645078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,8,balanced,0.02165866643190384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,8,balanced,0.02199999988079071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,8,balanced,0.0199946661790212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,8,balanced,0.02201066662867864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,8,balanced,0.024501333634058636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,8,balanced,0.042037333051363625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,8,balanced,0.04240000247955322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,8,balanced,0.038831998904546104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,8,balanced,0.03606933355331421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,8,balanced,0.03460799902677536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,8,power_law_1.01,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,8,balanced,0.044682666659355164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,8,power_law_1.01,0.06865919828414917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,8,balanced,0.04285866518815359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,8,power_law_1.01,0.06585599780082703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,8,balanced,0.036730666955312095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,8,power_law_1.01,0.07182080149650574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,8,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,8,power_law_1.01,0.06880000233650208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,8,balanced,0.04451199869314829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,8,power_law_1.01,0.06738560199737549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,8,power_law_1.01,0.06390399932861328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,8,balanced,0.046575998266537987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,8,power_law_1.01,0.06206079721450806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,8,balanced,0.05893866717815399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,8,power_law_1.01,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,8,power_law_1.01,0.05738239884376526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,8,balanced,0.06070933242638906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,8,power_law_1.01,0.04682239890098572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,8,power_law_1.01,0.0596671998500824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,8,power_law_1.01,0.04835839867591858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,8,balanced,0.08425066868464152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,8,power_law_1.01,0.055308800935745236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,8,power_law_1.01,0.05059199929237366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,8,balanced,0.0999840001265208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,8,power_law_1.01,0.05822719931602478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,8,power_law_1.01,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,8,power_law_1.01,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,8,power_law_1.01,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,8,balanced,0.13620266318321228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,8,power_law_1.01,0.0675711989402771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,8,power_law_1.01,0.04983679950237274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,8,power_law_1.01,0.07523840069770812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,8,power_law_1.01,0.046937599778175354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,8,power_law_1.01,0.08081279993057251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,8,power_law_1.01,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,8,power_law_1.01,0.08627200126647949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,8,power_law_1.01,0.044998401403427125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,8,power_law_1.01,0.04316799938678741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,8,power_law_1.01,0.11036159992218017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,8,power_law_1.01,0.05271040201187134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,8,power_law_1.01,0.1188480019569397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,8,power_law_1.01,0.04951040148735046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,8,power_law_1.01,0.14346879720687866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,8,power_law_1.01,0.05014399886131286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,8,power_law_1.01,0.18801920413970946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,8,power_law_1.01,0.05737599730491638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,8,power_law_1.01,0.2539328098297119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,8,power_law_1.01,0.06161919832229614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,8,power_law_1.01,0.33164799213409424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,8,power_law_1.01,0.06399359703063964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,8,power_law_1.01,0.4635136127471924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,8,power_law_1.01,0.08511999845504761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,8,power_law_1.01,0.6070784091949463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,8,power_law_1.01,0.04657920002937317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,8,power_law_1.01,0.09401599764823913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,8,power_law_1.01,0.8408703804016113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,8,power_law_1.01,0.046700799465179445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,8,power_law_1.01,0.12024320363998413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,8,power_law_1.01,1.1433088302612304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,8,power_law_1.01,0.04708479940891266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,8,power_law_1.01,0.1332159996032715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,8,power_law_1.01,0.0487744003534317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,8,power_law_1.01,1.5035584449768067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,8,power_law_1.01,0.20293118953704833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,8,power_law_1.01,0.04865919947624207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,8,power_law_1.01,2.361043167114258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,8,power_law_1.01,0.2472831964492798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,8,power_law_1.01,0.048172798752784726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,8,power_law_1.01,4.770880126953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,8,power_law_1.01,0.34570879936218263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,8,power_law_1.01,0.047731199860572816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,8,power_law_1.01,0.4402815818786621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,8,power_law_1.01,0.047200000286102294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,8,power_law_1.01,0.6419199943542481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,8,power_law_1.01,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,8,power_law_1.01,0.8796223640441895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,8,power_law_1.01,0.04609920084476471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,8,power_law_1.01,1.060159969329834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,8,power_law_1.01,0.04314239919185638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,8,power_law_1.01,0.046374401450157164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,8,power_law_1.01,1.6278335571289062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,8,power_law_1.01,0.05310720205307007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,8,power_law_1.01,3.2774913787841795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,8,power_law_1.01,0.04916479885578155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,8,power_law_1.01,0.05600000023841858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,8,power_law_1.01,0.056441599130630495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,8,power_law_1.01,0.06416640281677247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,8,power_law_1.01,0.07646719813346863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.01,0.08360959887504578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.01,0.101254403591156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.01,0.11962239742279053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.01,0.16072959899902345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.01,0.21018879413604735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.01,0.2892544031143188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.01,0.37482879161834715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.01,0.5521855831146241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.01,0.6688896179199219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,8,power_law_1.01,0.9102784156799316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,8,power_law_1.01,1.444422435760498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,8,power_law_1.01,2.619001579284668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,8,power_law_1.01,0.019174399971961974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,8,power_law_1.01,0.020172800123691558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,8,power_law_1.01,0.019424000382423402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,8,power_law_1.01,0.019916799664497376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,8,power_law_1.01,0.02077440023422241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,8,power_law_1.01,0.02306559979915619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,8,power_law_1.01,0.03997440040111542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,8,power_law_1.01,0.04012160003185272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,8,power_law_1.01,0.03656960129737854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,8,power_law_1.2,0.06727039813995361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,8,power_law_1.01,0.034585601091384886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,8,power_law_1.2,0.06858239769935608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,8,power_law_1.01,0.032953599095344545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,8,power_law_1.2,0.06546559929847717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,8,power_law_1.01,0.04312320053577423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,8,power_law_1.2,0.06775680184364319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,8,power_law_1.01,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,8,power_law_1.2,0.06769279837608337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,8,power_law_1.01,0.034720000624656674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,8,power_law_1.2,0.06598399877548218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,8,power_law_1.01,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,8,power_law_1.2,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,8,power_law_1.01,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,8,power_law_1.2,0.05914880037307739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,8,power_law_1.01,0.0456959992647171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,8,power_law_1.2,0.057036799192428586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,8,power_law_1.01,0.057580798864364624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,8,power_law_1.2,0.05583360195159912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.01,0.06045439839363098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,8,power_law_1.2,0.05487359762191772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.01,0.08412160277366638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,8,power_law_1.2,0.06540160179138184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.01,0.09968000054359435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,8,power_law_1.2,0.06474879980087281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.01,0.13591040372848512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,8,power_law_1.2,0.06625279784202576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.01,0.16965119838714598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,8,power_law_1.2,0.08314239978790283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,8,power_law_1.2,0.044300800561904906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.01,0.24396800994873047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,8,power_law_1.2,0.07780479788780212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,8,power_law_1.2,0.04826880097389221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.01,0.31607038974761964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,8,power_law_1.2,0.0864960014820099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,8,power_law_1.2,0.050809597969055174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.01,0.45348482131958007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,8,power_law_1.2,0.10609279870986939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,8,power_law_1.2,0.052767997980117796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,8,power_law_1.2,0.12191359996795655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.01,0.5918528079986572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,8,power_law_1.2,0.04912000000476837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,8,power_law_1.2,0.15482239723205565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,8,power_law_1.01,0.7341824054718018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,8,power_law_1.2,0.04806399941444397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,8,power_law_1.2,0.18861440420150757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,8,power_law_1.01,1.1530752182006836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,8,power_law_1.2,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,8,power_law_1.2,0.24636800289154054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,8,power_law_1.2,0.047219198942184445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,8,power_law_1.01,2.271743965148926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,8,power_law_1.2,0.33564159870147703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,8,power_law_1.2,0.047865599393844604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,8,balanced,0.16941332817077637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,8,power_law_1.2,0.4241983890533447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,8,power_law_1.2,0.04509440064430237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,8,power_law_1.2,0.6043456077575684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,8,power_law_1.2,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,8,balanced,0.24157865842183432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,8,power_law_1.2,0.04842880070209503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,8,power_law_1.2,0.8627712249755859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,8,power_law_1.2,0.050380802154541014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,8,power_law_1.2,1.1277055740356445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,8,balanced,0.31241599718729657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,8,power_law_1.2,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,8,power_law_1.2,1.4876352310180665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,8,power_law_1.2,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,8,power_law_1.2,2.287366485595703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,8,power_law_1.2,0.06454399824142457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,8,power_law_1.2,4.802630233764648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,8,power_law_1.2,0.06667519807815551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,8,balanced,0.4509119987487793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,8,power_law_1.2,0.07815679907798767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,8,power_law_1.2,0.08919680118560791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,8,power_law_1.2,0.11677440404891967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,8,balanced,0.592847983042399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,8,power_law_1.2,0.14431359767913818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,8,power_law_1.2,0.20238080024719238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,8,power_law_1.2,0.25080320835113523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,8,power_law_1.2,0.3605504035949707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,8,power_law_1.2,0.4302783966064453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,8,balanced,0.7356800238291422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,8,power_law_1.2,0.6545472145080566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,8,power_law_1.2,0.8390848159790039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,8,power_law_1.2,1.0904128074645996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,8,power_law_1.2,1.627903938293457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,8,power_law_1.2,3.4280574798583983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,8,balanced,1.1505813598632812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,8,balanced,2.271733283996582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,8,power_law_1.2,0.044787201285362246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,8,power_law_1.2,0.044761601090431216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,8,power_law_1.2,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,8,power_law_1.2,0.04690560102462769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,8,power_law_1.2,0.04756479859352112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,8,power_law_1.2,0.04738560020923614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,8,power_law_1.2,0.047705599665641786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,8,power_law_1.2,0.04756479859352112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,8,power_law_1.2,0.019449600577354433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,8,power_law_1.2,0.04618239998817444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,8,power_law_1.2,0.020947200059890748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,8,power_law_1.2,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,8,power_law_1.2,0.0205375999212265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,8,power_law_1.2,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,8,power_law_1.2,0.02024960070848465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,8,power_law_1.2,0.050783997774124144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,8,power_law_1.2,0.019897599518299103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,8,power_law_1.2,0.0515392005443573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,8,power_law_1.2,0.022944000363349915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,8,power_law_1.2,0.04913919866085052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,8,power_law_1.2,0.039647999405860904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,8,power_law_1.2,0.04002560079097748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,8,power_law_1.2,0.05914880037307739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,8,power_law_1.2,0.035692799091339114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,8,power_law_1.2,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,8,power_law_1.2,0.03468160033226013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,8,power_law_1.2,0.06328319907188415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,8,power_law_1.2,0.032691198587417605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,8,power_law_1.2,0.07211520075798035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,8,power_law_1.2,0.04323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.2,0.08446080088615418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,8,power_law_1.2,0.04037759900093078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.2,0.10330239534378052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,8,power_law_1.2,0.03513599932193756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.2,0.13140480518341063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,4,balanced,0.04818133513132731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,8,power_law_1.2,0.045388799905776975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.2,0.1745919942855835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,4,balanced,0.050570666790008545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,8,power_law_1.2,0.04353919923305512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,4,balanced,0.0589279979467392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.2,0.2189631938934326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,4,balanced,0.05528533458709717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,8,power_law_1.2,0.04478079974651337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.2,0.29183359146118165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,4,balanced,0.05766933163007101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,8,power_law_1.2,0.05772160291671753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.2,0.38417279720306396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,4,balanced,0.05555200080076853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.2,0.060089600086212155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.2,0.5292479991912842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,4,balanced,0.05429333448410034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.2,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.2,0.7469183921813964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,4,balanced,0.05526933570702871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.2,0.09916800260543823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,8,power_law_1.2,0.8468480110168457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,4,balanced,0.06046933432420095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,4,balanced,0.03997333347797394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.2,0.1390720009803772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,8,power_law_1.2,1.4332096099853515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,4,balanced,0.05495466788609823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,4,balanced,0.041002665956815086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.2,0.1701375961303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,4,balanced,0.05938666562239329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,8,power_law_1.2,2.8609088897705077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,4,balanced,0.04584000011285146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.2,0.2445375919342041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,4,balanced,0.05990933378537496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,4,balanced,0.04704533517360687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.2,0.315449595451355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,4,balanced,0.05247466762860616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,4,balanced,0.046298667788505554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.2,0.4542208194732666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,4,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,4,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.2,0.5959424018859864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,4,balanced,0.059343998630841575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,4,balanced,0.05282133320967356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,8,power_law_1.2,0.7358592033386231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,4,balanced,0.04685866832733154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,4,balanced,0.05486933390299479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,4,balanced,0.04633066554864248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,4,balanced,0.05894933144251505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,8,power_law_1.2,1.154310417175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,4,balanced,0.04728533327579498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,4,balanced,0.06597866614659627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,8,power_law_1.2,2.270854377746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,4,balanced,0.047194664676984154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,4,balanced,0.0692799985408783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,4,balanced,0.04880533119042715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,4,balanced,0.08760000268618266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,4,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,4,balanced,0.04831466575463613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,4,balanced,0.09195199608802795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,4,balanced,0.0509493350982666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,4,balanced,0.04882133503754934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,4,balanced,0.11615999539693196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,4,balanced,0.05314666529496511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,4,balanced,0.13131733735402426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,4,balanced,0.060405333836873375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,4,balanced,0.06260799864927928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,4,balanced,0.1787839929262797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,4,balanced,0.07783466577529907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,4,balanced,0.09358933568000793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,4,balanced,0.21210666497548422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,4,balanced,0.11961600184440613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,4,balanced,0.14086932937304178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,4,balanced,0.302837332089742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,4,balanced,0.18954133987426758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,4,balanced,0.2357706626256307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,4,balanced,0.38807467619578045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,4,balanced,0.33290666341781616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,4,balanced,0.427621324857076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,4,balanced,0.4740906556447347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,4,balanced,0.525984009106954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,4,balanced,0.7406240304311117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,4,balanced,0.8110293547312418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,4,balanced,1.4434773127237956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,4,balanced,1.5742613474527996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,4,balanced,0.04022933294375738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,4,balanced,0.042277331153551735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,4,balanced,0.04580800235271454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,4,balanced,0.04764266808827718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,4,balanced,0.047007997830708824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,4,balanced,0.04640000065167745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,4,balanced,0.048341333866119385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,4,balanced,0.049466664592425026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,4,balanced,0.01979200045267741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,4,balanced,0.048512001832326256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,4,balanced,0.020517333100239437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,4,balanced,0.05020266771316528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,4,balanced,0.020949333906173706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,4,balanced,0.046816001335779824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,4,balanced,0.02258133391539256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,4,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,4,balanced,0.02314666658639908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,4,balanced,0.04693866769472758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,4,balanced,0.04747200012207031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,4,balanced,0.052757332722345986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,4,balanced,0.04971733192602793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,4,balanced,0.04926399886608124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,4,balanced,0.06092800199985504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,4,balanced,0.06265600025653839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,4,balanced,0.06970133384068807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,4,balanced,0.08098133405049641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,4,balanced,0.09948800007502238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,4,balanced,0.12079466382662456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,4,balanced,0.1546880006790161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,4,balanced,0.18689066171646118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,4,balanced,0.2613973418871562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,4,balanced,0.33166933059692383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,4,balanced,0.40675731499989826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,4,balanced,0.614192008972168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,4,power_law_1.01,0.05289599895477295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,4,power_law_1.01,0.04028159976005554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,4,power_law_1.01,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,4,power_law_1.01,0.04561919867992401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,4,power_law_1.01,0.054176002740859985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,4,power_law_1.01,0.04083200097084046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,4,power_law_1.01,0.05623040199279785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,4,power_law_1.01,0.04679040014743805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,4,power_law_1.01,0.05775359869003296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,4,power_law_1.01,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,4,power_law_1.01,0.05937280058860779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,4,power_law_1.01,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,4,power_law_1.01,0.056524801254272464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,4,power_law_1.01,0.045440000295639035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,4,power_law_1.01,0.045977601408958436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,4,power_law_1.01,0.047040000557899475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,4,power_law_1.01,0.045798400044441225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,4,power_law_1.01,0.046214398741722104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,4,power_law_1.01,0.048179200291633605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,4,power_law_1.01,0.04683519899845123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,4,balanced,1.1794239679972331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,4,power_law_1.01,0.0502016007900238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,4,power_law_1.01,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,4,power_law_1.01,0.05247359871864319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,4,power_law_1.01,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,4,power_law_1.01,0.07425919771194459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,4,power_law_1.01,0.08760319948196411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,4,power_law_1.01,0.10786559581756591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,4,power_law_1.01,0.13661439418792726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,4,power_law_1.01,0.16391680240631104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,4,power_law_1.01,0.22301440238952636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,4,power_law_1.01,0.3245503902435303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,4,power_law_1.01,0.4306623935699463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,4,power_law_1.01,0.529753589630127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,4,power_law_1.01,0.7078656196594239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,4,power_law_1.01,0.9778495788574219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,4,power_law_1.01,1.5837632179260255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,4,power_law_1.01,3.423712158203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,4,balanced,0.021920000513394673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,4,balanced,0.02454400062561035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,4,balanced,0.024906667570273083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,4,balanced,0.04144000013669332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,4,power_law_1.01,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,4,balanced,0.04318933188915253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,4,power_law_1.01,0.04284160137176514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,4,balanced,0.0429066667954127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,4,power_law_1.01,0.04517120122909546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,4,balanced,0.03893866638342539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,4,power_law_1.01,0.04613119959831238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,4,balanced,0.03728533287843069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,4,power_law_1.01,0.045440000295639035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,4,balanced,0.03437866767247518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,4,power_law_1.01,0.04534400105476379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,4,balanced,0.038549333810806274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,4,power_law_1.01,0.04799999892711639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,4,balanced,0.036330667634805046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,4,power_law_1.01,0.04527359902858734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,4,balanced,0.03669333209594091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,4,power_law_1.01,0.04609920084476471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,4,balanced,0.048512001832326256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,4,power_law_1.01,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,4,balanced,0.05054399867852529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,4,power_law_1.01,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,4,balanced,0.06291733185450236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,4,power_law_1.01,0.0499455988407135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,4,balanced,0.06906666855017345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,4,power_law_1.01,0.01935359984636307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,4,power_law_1.01,0.01854719966650009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,4,balanced,0.09607999523480733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,4,power_law_1.01,0.018745599687099455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,4,power_law_1.01,0.019500799477100372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,4,balanced,0.11452266573905945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,4,power_law_1.01,0.020524799823760986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,4,power_law_1.01,0.02072319984436035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,4,balanced,0.15658133228619894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,4,power_law_1.01,0.023878400027751923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,4,power_law_1.01,0.023296000063419343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,4,power_law_1.01,0.040505599975585935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,4,power_law_1.01,0.04077439904212952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,4,balanced,0.19992534319559732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,4,power_law_1.01,0.04099839925765991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,4,power_law_1.01,0.03516159951686859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,4,power_law_1.01,0.034995201230049136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,4,power_law_1.01,0.0328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,4,balanced,0.2848320007324219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,4,power_law_1.01,0.037990400195121767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,4,power_law_1.01,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,4,power_law_1.01,0.03526400029659271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,4,power_law_1.01,0.04620159864425659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.01,0.048556798696517946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.01,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,4,balanced,0.36935468514760333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.01,0.07015680074691773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.01,0.09605759978294373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.01,0.11253759860992432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.01,0.1555840015411377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.01,0.19826560020446776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,4,balanced,0.4526453415552775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.01,0.2852992057800293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.01,0.3689471960067749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,4,power_law_1.01,0.4519167900085449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,4,power_law_1.01,0.7088128089904785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,4,power_law_1.01,1.3826496124267578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,4,balanced,0.7106453577677408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,4,power_law_1.01,0.05637120008468628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,4,power_law_1.01,0.056492799520492555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,4,power_law_1.01,0.056908798217773435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,4,power_law_1.01,0.05737599730491638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,4,power_law_1.01,0.05867519974708557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,4,power_law_1.01,0.06239359974861145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,4,power_law_1.01,0.0611840009689331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,4,power_law_1.01,0.06753919720649719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,4,power_law_1.01,0.0720255970954895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,4,power_law_1.01,0.07616000175476074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,4,power_law_1.01,0.08739839792251587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,4,balanced,1.3843305905659993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,4,power_law_1.01,0.10089600086212158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,4,power_law_1.01,0.117958402633667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,4,power_law_1.01,0.16395519971847533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,4,power_law_1.01,0.21578240394592285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,4,power_law_1.01,0.23484160900115966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,4,power_law_1.01,0.4096127986907959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,4,power_law_1.01,0.4687039852142334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,4,power_law_1.01,0.6172095775604248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,4,power_law_1.01,0.7322495937347412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,4,power_law_1.01,1.144863986968994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,4,power_law_1.01,1.5595647811889648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,4,power_law_1.01,4.087353515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,4,power_law_1.2,0.05994880199432373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,4,power_law_1.2,0.04943360090255737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,4,power_law_1.2,0.05338240265846252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,4,power_law_1.2,0.055430400371551516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,4,power_law_1.2,0.057081598043441775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,4,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,4,power_law_1.2,0.05454720258712768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,4,power_law_1.2,0.056985598802566526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,4,power_law_1.2,0.05717759728431702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,4,power_law_1.2,0.05616639852523804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,4,power_law_1.2,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,4,power_law_1.2,0.057843202352523805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,4,power_law_1.2,0.061222398281097413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,4,power_law_1.2,0.06207360029220581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,4,power_law_1.2,0.06381440162658691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,4,power_law_1.2,0.0713919997215271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,4,power_law_1.2,0.042182400822639465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,4,power_law_1.2,0.0785215973854065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,4,power_law_1.2,0.04348799884319306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,4,power_law_1.2,0.09612159729003907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,4,power_law_1.2,0.09567360281944275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,4,power_law_1.2,0.039417600631713866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,4,power_law_1.2,0.12528640031814575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,4,power_law_1.2,0.04442239999771118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,4,power_law_1.2,0.155731201171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,4,power_law_1.2,0.04465279877185822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,4,power_law_1.2,0.19553279876708984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,4,power_law_1.2,0.04711039960384369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,4,power_law_1.2,0.23781120777130127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,4,power_law_1.2,0.047328001260757445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,4,power_law_1.2,0.04588800072669983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,4,power_law_1.2,0.35196800231933595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,4,power_law_1.2,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,4,power_law_1.2,0.042668798565864564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,4,power_law_1.01,0.04943360090255737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,4,power_law_1.2,0.5397823810577392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,4,power_law_1.2,0.04622719883918762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,4,power_law_1.2,0.042156800627708435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,4,power_law_1.01,0.05140479803085327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,4,power_law_1.2,0.0466623991727829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,4,power_law_1.2,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,4,power_law_1.2,0.8839103698730468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,4,power_law_1.01,0.056959998607635495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,4,power_law_1.2,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,4,power_law_1.2,0.0474047988653183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,4,power_law_1.2,0.9934528350830079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,4,power_law_1.01,0.0585919976234436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,4,power_law_1.2,0.04865919947624207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,4,power_law_1.2,0.04570879936218262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,4,power_law_1.2,1.1355135917663575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,4,power_law_1.01,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,4,power_law_1.2,0.04876160025596619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,4,power_law_1.2,0.045977601408958436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,4,power_law_1.01,0.06753919720649719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,4,power_law_1.2,1.7747711181640624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,4,power_law_1.2,0.048172798752784726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,4,power_law_1.2,0.046649599075317384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.01,0.07953919768333435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,4,power_law_1.2,3.5979713439941405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,4,power_law_1.2,0.05495679974555969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,4,power_law_1.2,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.01,0.09568639993667602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,4,power_law_1.2,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,4,power_law_1.2,0.04671359956264496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.01,0.10792319774627686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,4,power_law_1.2,0.06586880087852479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.01,0.16388479471206666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,4,power_law_1.2,0.07857279777526856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.01,0.1720128059387207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,4,power_law_1.2,0.08670719861984252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.01,0.25148160457611085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,4,power_law_1.2,0.11012480258941651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.01,0.35211520195007323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,4,power_law_1.2,0.13489919900894165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.01,0.5259071826934815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,4,power_law_1.2,0.1933632016181946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.01,0.719539213180542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,4,power_law_1.2,0.22108800411224366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,4,power_law_1.01,0.7879168033599854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,4,power_law_1.2,0.3357248067855835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,4,power_law_1.01,1.1270208358764648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,4,power_law_1.2,0.44423680305480956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,4,power_law_1.01,2.4984767913818358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,4,power_law_1.2,0.6419648170471192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,4,power_law_1.2,0.800972843170166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,4,power_law_1.2,1.181824016571045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,4,power_law_1.2,1.8883455276489258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,4,power_law_1.2,3.0519807815551756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,4,power_law_1.2,0.018329599499702455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,4,power_law_1.2,0.018873600661754607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,4,power_law_1.2,0.0190528005361557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,4,power_law_1.2,0.018771199882030486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,4,power_law_1.2,0.019974400103092194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,4,power_law_1.2,0.020703999698162077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,4,power_law_1.2,0.02398719936609268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,4,power_law_1.2,0.023123200237751006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,2,balanced,0.04035199930270513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,4,power_law_1.2,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,2,balanced,0.042821332812309265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,4,power_law_1.2,0.03999359905719757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,2,balanced,0.05030933519204458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,4,power_law_1.2,0.040659201145172116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,2,balanced,0.055530667304992676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,4,power_law_1.2,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,2,balanced,0.05045866469542185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,4,power_law_1.2,0.0348800003528595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,2,balanced,0.0528053343296051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,4,power_law_1.2,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,2,balanced,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,4,power_law_1.2,0.03798399865627289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,2,balanced,0.0529120018084844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,4,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,2,balanced,0.059706668059031166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,4,power_law_1.2,0.036883199214935304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,2,balanced,0.036303999523321785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,2,balanced,0.05458133419354757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,4,power_law_1.2,0.046828800439834596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,2,balanced,0.04127466678619385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,2,balanced,0.05182399849096934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.2,0.04987519979476929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,2,balanced,0.04268800218900045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,2,balanced,0.05339199801286062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,2,balanced,0.04683733483155569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.2,0.06311039924621582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,2,balanced,0.05291733145713806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,2,balanced,0.04667200148105621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.2,0.07130240201950074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,2,balanced,0.053173333406448364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,2,balanced,0.045066664616266884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.2,0.09705600142478943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,2,balanced,0.055760001142819725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,2,balanced,0.04950400193532308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.2,0.11336959600448608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,2,balanced,0.055311997731526695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,2,balanced,0.0452106644709905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.2,0.1563647985458374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,2,balanced,0.059077332417170204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,2,balanced,0.04941866795221964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.2,0.19985920190811157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,2,balanced,0.0714026689529419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,2,balanced,0.04541866481304169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.2,0.28178560733795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,2,balanced,0.07374399900436401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,2,balanced,0.04670399924119314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.2,0.36910719871520997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,2,balanced,0.048783997694651283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,2,balanced,0.09200533231099446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,4,power_law_1.2,0.4544960021972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,2,balanced,0.04532266656557719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,2,balanced,0.102101335922877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,2,balanced,0.04631466666857401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,4,power_law_1.2,0.7080063819885254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,2,balanced,0.05115200082461039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,4,power_law_1.2,1.3757247924804688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,2,balanced,0.1291093329588572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,2,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,2,balanced,0.05148266752560934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,2,balanced,0.14412267009417215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,2,balanced,0.061936000982920326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,2,balanced,0.06939200063546498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,2,balanced,0.19896533091862997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,2,balanced,0.08804800113042195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,2,balanced,0.10538132985432942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,2,balanced,0.24410667022069296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,2,balanced,0.13336533308029175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,2,balanced,0.34750401973724365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,2,balanced,0.1594986617565155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,2,balanced,0.22245866060256958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,2,balanced,0.45049599806467694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,2,balanced,0.2760159969329834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,4,power_law_1.2,0.04490239918231964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,2,balanced,0.39773865540822345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,2,balanced,0.5506240129470825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,4,power_law_1.2,0.0486272007226944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,4,power_law_1.2,0.05149440169334411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,4,power_law_1.2,0.05125759840011597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,4,power_law_1.2,0.056979197263717654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,2,balanced,0.515343983968099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,4,power_law_1.2,0.060127997398376466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,4,power_law_1.2,0.0636031985282898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,2,balanced,0.8546026547749838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,4,power_law_1.2,0.06922240257263183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.2,0.07964159846305847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.2,0.0908352017402649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.2,0.1227903962135315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,2,balanced,0.6353546778361002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.2,0.17811199426651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.2,0.21658239364624024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.2,0.2950527906417847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.2,0.3788032054901123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.2,0.4947519779205322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.2,0.6264959812164307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,4,power_law_1.2,0.9957568168640136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,4,power_law_1.2,1.493619155883789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,2,balanced,0.9857119719187418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,2,balanced,1.6808212598164876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,4,power_law_1.2,2.3389055252075197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,2,balanced,1.9164427121480305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,2,balanced,0.03585600107908249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,2,balanced,0.04353066782156626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,2,balanced,0.04693866769472758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,2,balanced,0.05096533397833506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,2,balanced,0.045040001471837364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,2,balanced,0.0491839994986852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,2,balanced,0.04949333270390829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,2,balanced,0.04722133278846741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,2,balanced,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,2,balanced,0.04876266419887543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,2,balanced,0.048394665122032166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,2,balanced,0.04879466692606608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,2,balanced,0.047370667258898415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,2,balanced,0.04884799818197886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,2,balanced,0.05338666836420695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,2,balanced,0.05273066461086273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,2,balanced,0.05284800132115682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,2,balanced,0.06490666667620341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,2,balanced,0.06924800078074138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,2,balanced,0.07838400204976399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,2,balanced,0.09167466560999553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,2,balanced,0.11896000305811565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,2,balanced,0.14446933070818582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,2,balanced,0.1916853388150533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,2,balanced,0.2395306626955668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,2,balanced,0.33947734038035077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,2,balanced,0.4342720111211141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,2,balanced,0.019333332777023315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,2,balanced,0.019754666835069656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,2,balanced,0.019882666567961376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,2,balanced,0.0204373337328434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,2,balanced,0.022128000855445862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,2,balanced,0.5323306719462076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,2,balanced,0.02205866575241089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,2,balanced,0.0220266655087471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,2,balanced,0.022522665560245514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,2,balanced,0.024319998919963837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,2,power_law_1.01,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,2,balanced,0.024282666544119518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,2,power_law_1.01,0.046163201332092285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,2,balanced,0.024682665864626568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,2,balanced,0.8304800192515055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,2,power_law_1.01,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,2,balanced,0.04246933261553446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,2,power_law_1.01,0.05307520031929016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,2,balanced,0.04274133344491323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,2,power_law_1.01,0.05516160130500793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,2,balanced,0.04278933505217234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,2,power_law_1.01,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,2,balanced,0.03416533271471659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,2,power_law_1.01,0.05422719717025757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,2,balanced,0.03458133339881897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,2,power_law_1.01,0.05671679973602295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,2,balanced,0.036229332288106285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,2,power_law_1.01,0.0382207989692688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,2,power_law_1.01,0.05437440276145935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,2,balanced,0.0423573354880015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,2,power_law_1.01,0.0551360011100769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,2,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,2,power_law_1.01,0.054579198360443115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,2,balanced,0.05509333312511444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,2,power_law_1.01,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,2,power_law_1.01,0.05811840295791626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,2,balanced,0.061146666606267296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,2,balanced,1.5992639859517415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,2,power_law_1.01,0.05959039926528931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,2,power_law_1.01,0.06514559984207154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,2,balanced,0.07981333136558533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,2,power_law_1.01,0.06724479794502258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,2,balanced,0.08919466535250346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,2,power_law_1.01,0.07079039812088013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,2,power_law_1.01,0.08339840173721313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,2,power_law_1.01,0.08906880021095276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,2,balanced,0.12473066647847493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,2,power_law_1.01,0.11184639930725097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,2,power_law_1.01,0.13217920064926147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,2,balanced,0.1511146624883016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,2,power_law_1.01,0.15673600435256957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,2,power_law_1.01,0.20734720230102538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,2,power_law_1.01,0.27550079822540285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,2,power_law_1.01,0.33882238864898684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,2,balanced,0.2127466599146525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,2,power_law_1.01,0.5046783924102783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,2,power_law_1.01,0.6272128105163575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,2,power_law_1.01,0.8081088066101074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,2,balanced,0.2728479901949565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,2,power_law_1.01,1.2624064445495606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,2,power_law_1.01,2.764723205566406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,2,balanced,0.33475200335184735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,2,balanced,0.5201866626739502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,2,balanced,1.0058666865030925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,2,power_law_1.01,0.0392767995595932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,2,power_law_1.01,0.039340800046920775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,2,power_law_1.01,0.041766399145126344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,2,power_law_1.01,0.04864639937877655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,2,power_law_1.01,0.048607999086380006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,2,power_law_1.01,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,2,power_law_1.01,0.04730879962444305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,2,power_law_1.01,0.0473471999168396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,2,power_law_1.01,0.04845440089702606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,2,power_law_1.01,0.047635200619697574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,2,power_law_1.01,0.047884801030159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,2,power_law_1.01,0.04838399887084961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,2,power_law_1.01,0.050329601764678954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,2,power_law_1.01,0.05050240159034729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,2,power_law_1.01,0.05730559825897217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,2,power_law_1.01,0.0582144021987915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,2,power_law_1.01,0.017888000607490538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,2,power_law_1.01,0.06330239772796631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,2,power_law_1.01,0.01860480010509491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,2,power_law_1.01,0.07326080203056336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,2,power_law_1.01,0.018572799861431122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.01,0.0805184006690979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.01,0.09646720290184022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,2,power_law_1.01,0.01860480010509491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.01,0.12347520589828491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,2,power_law_1.01,0.019705599546432494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.01,0.14104959964752198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,2,power_law_1.01,0.019577600061893463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,2,power_law_1.01,0.020947200059890748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.01,0.1769919991493225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,2,power_law_1.01,0.021004800498485566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.01,0.26044158935546874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,2,power_law_1.01,0.02311040014028549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.01,0.349785590171814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,2,power_law_1.01,0.02276480048894882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.01,0.43471360206604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,2,power_law_1.01,0.023552000522613525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.01,0.5975039958953857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,2,power_law_1.01,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,2,power_law_1.01,0.8230079650878906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,2,power_law_1.01,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,2,power_law_1.01,1.1240511894226075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,2,power_law_1.01,0.041529598832130435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,2,power_law_1.01,2.4493696212768556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,2,power_law_1.01,0.034720000624656674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,2,power_law_1.01,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,2,power_law_1.01,0.03523840010166168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,2,power_law_1.01,0.03826560080051422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.01,0.042822399735450746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.01,0.053939199447631835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.01,0.06054400205612183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.01,0.07843199968338013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.01,0.08880000114440918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.01,0.12266240119934083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.01,0.14948480129241942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,2,power_law_1.2,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.01,0.21288959980010985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,2,power_law_1.2,0.046291199326515195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.01,0.27254400253295896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,2,power_law_1.2,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,2,power_law_1.01,0.3347520112991333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,2,power_law_1.2,0.055718398094177245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,2,power_law_1.01,0.519052791595459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,2,power_law_1.2,0.05319039821624756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,2,power_law_1.01,1.005401611328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,2,power_law_1.2,0.05608320236206055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,2,power_law_1.2,0.054790401458740236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,2,power_law_1.01,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,2,power_law_1.2,0.05384960174560547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,2,power_law_1.01,0.03985919952392578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,2,power_law_1.2,0.05286399722099304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,2,power_law_1.01,0.044614401459693906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,2,power_law_1.2,0.05358719825744629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,2,power_law_1.01,0.04423680007457733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,2,power_law_1.2,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,2,power_law_1.01,0.04631040096282959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,2,power_law_1.2,0.05785599946975708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,2,power_law_1.01,0.044819200038909913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,2,power_law_1.2,0.059724801778793336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,2,power_law_1.01,0.04459519982337952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,2,power_law_1.2,0.05839999914169312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,2,power_law_1.01,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,2,power_law_1.01,0.045535999536514285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,2,power_law_1.2,0.06483839750289917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,2,power_law_1.01,0.044972801208496095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,2,power_law_1.2,0.06579840183258057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,2,power_law_1.2,0.07141119837760926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,2,power_law_1.01,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,2,power_law_1.01,0.04725759923458099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,2,power_law_1.2,0.08282880187034607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,2,power_law_1.01,0.04888319969177246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,2,power_law_1.2,0.09210240244865417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,2,power_law_1.01,0.05373439788818359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,2,power_law_1.2,0.11429760456085206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,2,power_law_1.01,0.06024320125579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,2,power_law_1.2,0.13697919845581055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,2,power_law_1.01,0.06106240153312683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,2,power_law_1.2,0.1815616011619568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,2,power_law_1.01,0.07504640221595764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,2,power_law_1.2,0.22288639545440675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,2,power_law_1.01,0.0847487986087799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,2,power_law_1.2,0.29697279930114745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,2,power_law_1.01,0.1019711971282959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,2,power_law_1.2,0.3502016067504883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,2,power_law_1.01,0.12175999879837036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,2,power_law_1.2,0.5633664131164551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,2,power_law_1.01,0.1653887987136841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,2,power_law_1.2,0.6791232109069825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,2,power_law_1.01,0.19420160055160524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,2,power_law_1.2,0.8372032165527343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,2,power_law_1.01,0.30271360874176023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,2,power_law_1.2,1.4650431632995606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,2,power_law_1.01,0.3437887907028198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,2,power_law_1.2,2.755283164978027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,2,power_law_1.01,0.522380781173706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,2,power_law_1.01,0.6533567905426025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,2,power_law_1.01,0.7988416194915772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,2,power_law_1.01,1.3449600219726563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,2,power_law_1.01,2.6279232025146486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,2,power_law_1.2,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,2,power_law_1.2,0.03945600092411041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,2,power_law_1.2,0.037049600481987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,2,power_law_1.2,0.04413439929485321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,2,power_law_1.2,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,2,power_law_1.2,0.04462080001831055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,2,power_law_1.2,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,2,power_law_1.2,0.04533120095729828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,2,power_law_1.2,0.04514560103416443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,2,power_law_1.2,0.045952001214027406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,2,power_law_1.2,0.0455487996339798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,2,power_law_1.2,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,2,power_law_1.2,0.04860160052776337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,2,power_law_1.2,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,2,power_law_1.2,0.056006401777267456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,2,power_law_1.2,0.057715201377868654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,2,power_law_1.2,0.059680002927780154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,2,power_law_1.2,0.07399680018424988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,2,power_law_1.2,0.08332160115242004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,2,power_law_1.2,0.10524159669876099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,2,power_law_1.2,0.037945601344108584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,2,power_law_1.2,0.12438399791717529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,2,power_law_1.2,0.04181120097637177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,2,power_law_1.2,0.16526080369949342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,2,power_law_1.2,0.04289920032024384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,2,power_law_1.2,0.19887360334396362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,2,power_law_1.2,0.0489984005689621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,2,power_law_1.2,0.28748159408569335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,2,power_law_1.2,0.04708479940891266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,2,power_law_1.2,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,2,power_law_1.2,0.3888447999954224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,2,power_law_1.2,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,2,power_law_1.2,0.5489664077758789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,2,power_law_1.2,0.04584319889545441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,2,power_law_1.2,0.7132287979125976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,2,power_law_1.2,0.04766719937324524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,2,power_law_1.2,0.8020159721374511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,2,power_law_1.2,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,2,power_law_1.2,1.3118144035339356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,2,power_law_1.2,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,2,power_law_1.2,2.730284881591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,2,power_law_1.2,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,2,power_law_1.2,0.05115519762039185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,2,power_law_1.2,0.04978559911251068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,2,power_law_1.2,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,2,power_law_1.2,0.059059202671051025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,2,power_law_1.2,0.06309120059013366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,2,power_law_1.2,0.06881920099258423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.2,0.07946879863739013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.2,0.10261759757995606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.2,0.11261440515518188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.2,0.14451839923858642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,2,power_law_1.2,0.018297599256038667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.2,0.18173439502716066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,2,power_law_1.2,0.018636800348758698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.2,0.25231359004974363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,2,power_law_1.2,0.01847040057182312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.2,0.3247488021850586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,2,power_law_1.2,0.019731199741363524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.2,0.46800642013549804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,2,power_law_1.2,0.020467199385166168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.2,0.6385216236114502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,2,power_law_1.2,0.020179200172424316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,2,power_law_1.2,0.8394559860229492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,2,power_law_1.2,0.020364800095558168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,2,power_law_1.2,0.021452799439430237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,2,power_law_1.2,1.2712320327758788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,2,power_law_1.2,0.02253440022468567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,2,power_law_1.2,2.2579839706420897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,1,balanced,0.04241600135962168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,2,power_law_1.2,0.02449920028448105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,1,balanced,0.04574400186538696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,2,power_law_1.2,0.024217599630355836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,1,balanced,0.053786665201187134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,2,power_law_1.2,0.039673599600791934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,1,balanced,0.051541333397229515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,2,power_law_1.2,0.042508798837661746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,1,balanced,0.0509493350982666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,2,power_law_1.2,0.04418559968471527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,1,balanced,0.05333866675694784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,2,power_law_1.2,0.03407360017299652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,1,balanced,0.051967998345692955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,2,power_law_1.2,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,1,balanced,0.05332799752553304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,2,power_law_1.2,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,1,balanced,0.05345066885153452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,2,power_law_1.2,0.0409280002117157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,1,balanced,0.05341866612434387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.2,0.0414463996887207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,1,balanced,0.052970667680104576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.2,0.053414398431777955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.2,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.2,0.0784063994884491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.2,0.08867200016975403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.2,0.12387839555740357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.2,0.14975359439849853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.2,0.2121664047241211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.2,0.27325439453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,2,power_law_1.2,0.3343424081802368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,2,power_law_1.2,0.5197696208953857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,2,power_law_1.2,1.003865623474121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,1,balanced,0.036144000788529716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,1,balanced,0.04045866678158442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,1,balanced,0.04345066845417023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,1,balanced,0.046911999583244324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,1,balanced,0.04553600152333578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,1,balanced,0.045909335215886436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,1,balanced,0.047151997685432434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,1,balanced,0.047093331813812256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,1,balanced,0.04749333361784617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,1,balanced,0.04671466847260793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,1,balanced,0.046906664967536926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,1,balanced,0.04911999901135763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,1,balanced,0.0487413356701533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,1,balanced,0.04913066824277242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,1,balanced,0.05649066468079885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,1,balanced,0.055311997731526695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,1,balanced,0.05893866717815399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,1,balanced,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,1,balanced,0.08313066760698955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,1,balanced,0.10752000411351521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,1,balanced,0.1278986632823944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,1,power_law_1.01,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,1,balanced,0.16579733292261759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,1,power_law_1.01,0.041465601325035094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,1,power_law_1.01,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,1,balanced,0.2039733330408732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,1,power_law_1.01,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,1,power_law_1.01,0.052249598503112796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,1,balanced,0.2908053398132324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,1,power_law_1.01,0.05187839865684509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,1,power_law_1.01,0.05355520248413086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,1,power_law_1.01,0.0535103976726532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,1,balanced,0.36902932325998944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,1,power_law_1.01,0.05414400100708008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,1,power_law_1.01,0.053600001335144046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,1,power_law_1.01,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,1,power_law_1.01,0.05470079779624939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,1,balanced,0.5375466744105021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,1,power_law_1.01,0.0568448007106781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,1,power_law_1.01,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,1,power_law_1.01,0.06519039869308471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,1,power_law_1.01,0.0659712016582489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,1,power_law_1.01,0.06899840235710145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,1,balanced,0.7020746866861979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,1,power_law_1.01,0.07679359912872315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,1,power_law_1.01,0.08691200017929077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,1,power_law_1.01,0.10710400342941284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,1,power_law_1.01,0.11811200380325318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,1,power_law_1.01,0.15495680570602416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,1,power_law_1.01,0.19084800481796266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,1,balanced,0.8677919705708822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,1,power_law_1.01,0.25690879821777346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,1,power_law_1.01,0.3251840114593506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,1,power_law_1.01,0.45472002029418945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,1,power_law_1.01,0.5883264064788818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,1,power_law_1.01,0.035129600763320924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,1,power_law_1.01,0.7210815906524658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,1,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,1,power_law_1.01,1.1227392196655273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,1,power_law_1.01,0.03880960047245026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,1,power_law_1.01,0.04465279877185822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,1,power_law_1.01,2.197529602050781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,1,balanced,1.355247974395752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,1,power_law_1.01,0.04528000056743622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,1,power_law_1.01,0.04796159863471985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,1,power_law_1.01,0.044710400700569156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,1,power_law_1.01,0.046777600049972536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,1,power_law_1.01,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,1,power_law_1.01,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,1,power_law_1.01,0.04714879989624023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,1,power_law_1.01,0.04801279902458191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,1,power_law_1.01,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,1,power_law_1.01,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,1,power_law_1.01,0.0567359983921051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,1,power_law_1.01,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,1,power_law_1.01,0.06273919939994813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,1,power_law_1.01,0.07463679909706115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,1,balanced,2.6533066431681314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,1,power_law_1.01,0.08891519904136658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,1,power_law_1.01,0.11000319719314575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,1,power_law_1.01,0.12761600017547609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,1,power_law_1.01,0.1666815996170044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,1,power_law_1.01,0.20967040061950684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,1,power_law_1.01,0.29120640754699706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,1,power_law_1.01,0.374617600440979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,1,power_law_1.01,0.5365952014923095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,1,power_law_1.01,0.7061567783355713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,1,power_law_1.01,0.8636223793029785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,1,power_law_1.01,1.3522111892700195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,1,power_law_1.01,2.6503679275512697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,1,balanced,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,1,balanced,0.05450133482615153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,1,balanced,0.056159997979799904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,1,balanced,0.06393600006898244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,1,balanced,0.062458669145902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,1,balanced,0.06696000198523204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,1,balanced,0.0774239997069041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,1,balanced,0.08493333061536153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,1,balanced,0.10586133599281311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,1,balanced,0.11873066425323486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,1,balanced,0.15466666221618652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,1,balanced,0.18092266718546549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,1,balanced,0.25379733244578045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,1,balanced,0.31561599175135296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,1,balanced,0.45114131768544513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,1,balanced,0.5871306657791138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,1,power_law_1.2,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,1,balanced,0.720298687616984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,1,power_law_1.2,0.04177280068397522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,1,power_law_1.2,0.04226559996604919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,1,power_law_1.2,0.05182719826698303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,1,power_law_1.2,0.052774399518966675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,1,power_law_1.2,0.05075200200080872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,1,power_law_1.2,0.052153599262237546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,1,power_law_1.2,0.05122560262680054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,1,balanced,1.1198879877726238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,1,power_law_1.2,0.052774399518966675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,1,power_law_1.2,0.05367040038108826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,1,power_law_1.2,0.0525439977645874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,1,power_law_1.2,0.05395200252532959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,1,power_law_1.2,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,1,power_law_1.2,0.05748479962348938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,1,power_law_1.2,0.06340479850769043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,1,power_law_1.2,0.03528960049152374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,1,power_law_1.2,0.06541439890861511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,1,power_law_1.2,0.03699840009212494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,1,power_law_1.2,0.06895359754562377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,1,power_law_1.2,0.03859840035438537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,1,power_law_1.2,0.0773311972618103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,1,power_law_1.2,0.045049598813056944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,1,power_law_1.2,0.08635519742965699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,1,power_law_1.2,0.043961599469184875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,1,power_law_1.2,0.10803200006484985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,1,power_law_1.2,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,1,power_law_1.2,0.11896320581436157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,1,balanced,2.1947946548461914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,1,power_law_1.2,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,1,power_law_1.2,0.1529088020324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,1,power_law_1.2,0.046515199542045596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,1,power_law_1.2,0.19242240190505983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,1,power_law_1.2,0.04649600088596344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,1,power_law_1.2,0.2565119981765747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,1,power_law_1.2,0.04542720019817352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,1,power_law_1.2,0.3267199993133545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,1,power_law_1.2,0.04666880071163178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,1,power_law_1.2,0.45842561721801756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,1,power_law_1.2,0.04720639884471893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,1,power_law_1.2,0.5871488094329834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,1,power_law_1.2,0.049472001194953916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,1,power_law_1.2,0.7193984031677246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,1,power_law_1.2,0.04899199903011322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,1,power_law_1.2,1.1136639595031739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,1,power_law_1.2,0.05723519921302796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,1,power_law_1.2,0.05989760160446167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,1,power_law_1.2,2.195033645629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,8,balanced,0.09130133191744487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,1,power_law_1.2,0.06490240097045899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,8,balanced,0.09962667028109233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,1,power_law_1.2,0.07565439939498901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,8,balanced,0.09861866633097331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,1,power_law_1.2,0.08773120045661927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,8,balanced,0.1027733286221822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,1,power_law_1.2,0.1109503984451294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,8,balanced,0.09419199824333191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,1,power_law_1.2,0.12698240280151368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,8,balanced,0.09964266419410706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,1,power_law_1.2,0.16711039543151857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,8,balanced,0.09826667110125224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,1,power_law_1.2,0.21056001186370848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,8,balanced,0.092357337474823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,1,power_law_1.2,0.289849591255188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,8,balanced,0.09802666306495667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,8,balanced,0.094458669424057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,1,power_law_1.2,0.3742271900177002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,8,balanced,0.09447999795277913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,1,power_law_1.2,0.5418240070343018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,8,balanced,0.08799466490745544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,1,power_law_1.2,0.7025472164154053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,8,balanced,0.0848640004793803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,1,power_law_1.2,0.8665216445922852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,8,balanced,0.08295999964078267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,1,power_law_1.2,1.3517696380615234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,8,balanced,0.08235733211040497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,1,power_law_1.2,2.6526655197143554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,8,balanced,0.07977066437403361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,8,balanced,0.08401067058245341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,8,balanced,0.09315199653307597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,8,balanced,0.09030933181444804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,8,balanced,0.11371733744939168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,8,balanced,0.13170666495958963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,8,balanced,0.16684800386428833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,8,balanced,0.20585066080093384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,8,balanced,0.2720320026079814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,8,balanced,0.3429226477940877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,8,balanced,0.4835999806722005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,8,balanced,0.6246506770451864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,8,balanced,0.7484800020853678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,8,balanced,1.1531360149383545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,8,balanced,0.06057600180308024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,8,balanced,0.06097066899140676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,8,balanced,2.230384031931559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,8,balanced,0.06133866806825002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,8,balanced,0.06547733147939046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,8,balanced,0.06529066463311513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,8,balanced,0.06408533453941345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,8,balanced,0.058677335580190025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,8,balanced,0.06358399987220764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,8,balanced,0.05932799975077311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,8,balanced,0.06363200147946675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,8,balanced,0.05881066620349884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,8,balanced,0.06504000226656596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,8,balanced,0.0588266650835673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,8,balanced,0.062447999914487205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,8,balanced,0.059232001503308616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,8,balanced,0.06400000055631001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,8,balanced,0.05771199862162272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,8,balanced,0.06177600224812826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,8,balanced,0.05770133435726166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,8,balanced,0.06308266520500183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,8,balanced,0.05751466751098633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,8,balanced,0.05989866455396017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,8,balanced,0.059248000383377075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,8,balanced,0.060922667384147644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,8,balanced,0.056890666484832764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,8,balanced,0.06193066636721293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,8,balanced,0.057061334451039634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,8,balanced,0.05681600173314413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,8,balanced,0.06311999758084615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,8,balanced,0.056645333766937256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,8,balanced,0.07971199850241344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,8,balanced,0.05497066676616669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,8,balanced,0.0813920001188914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,8,balanced,0.054272000988324486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,8,balanced,0.10398933291435242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,8,balanced,0.056143999099731445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,8,balanced,0.1195093293984731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,8,balanced,0.05423999826113383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,8,balanced,0.07266666491826375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,8,balanced,0.15079999963442484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,8,balanced,0.02327466756105423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,8,balanced,0.06811200082302094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,8,balanced,0.024906667570273083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,8,balanced,0.024149333437283833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,8,balanced,0.18467734257380167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,8,balanced,0.08328533172607422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,8,balanced,0.023557332654794056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,8,balanced,0.09583466251691182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,8,balanced,0.02402666707833608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,8,balanced,0.2531200051307678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,8,balanced,0.026901334524154663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,8,balanced,0.12225066622098286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,8,balanced,0.049498667319615684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,8,balanced,0.05085866649945577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,8,balanced,0.31498666604359943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,8,balanced,0.1434986690680186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,8,balanced,0.047040000557899475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,8,balanced,0.043061330914497375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,8,balanced,0.19007466236750284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,8,balanced,0.03913066784540812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,8,balanced,0.05356800059477488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,8,balanced,0.4480266571044922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,8,balanced,0.05352533360322317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,8,balanced,0.23425600926081339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,8,balanced,0.04131733377774557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,8,balanced,0.06746133168538411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,8,balanced,0.054005334774653115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,8,balanced,0.5782506863276163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,8,balanced,0.32447999715805054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,8,balanced,0.061306665341059365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,8,balanced,0.07372800012429555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,8,balanced,0.09030399719874065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,8,balanced,0.12312533458073933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,8,balanced,0.41595200697580975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,8,balanced,0.7030133406321207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,8,balanced,0.14871999621391296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,8,balanced,0.20668800671895346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,8,balanced,0.25776533285776776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,8,balanced,0.50273064772288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,8,balanced,0.3672320048014323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,8,balanced,1.1009493668874104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,8,balanced,0.4750293493270874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,8,balanced,0.7733866373697916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,8,balanced,0.6947999795277914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,8,balanced,0.9115413029988607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,8,balanced,2.1538613637288413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,8,balanced,1.1292906602223713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,8,balanced,1.5088426272074382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,8,balanced,1.7785439491271973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,8,power_law_1.01,0.08858879804611205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,8,power_law_1.01,0.0917631983757019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,8,power_law_1.01,0.09687039852142335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,8,power_law_1.01,0.09393280148506164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,8,power_law_1.01,0.09375360012054443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,8,power_law_1.01,0.09159680008888245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,8,power_law_1.01,0.09268479943275451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,8,power_law_1.01,0.08945919871330262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,8,power_law_1.01,0.08107519745826722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,8,power_law_1.01,0.08079360127449035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,8,power_law_1.01,0.07856640219688416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,8,power_law_1.01,0.0934719979763031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,8,power_law_1.01,0.0869055986404419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,8,power_law_1.01,0.08930559754371643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,8,balanced,3.5154825846354165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,8,power_law_1.01,0.11240320205688477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,8,power_law_1.01,0.1137984037399292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,8,power_law_1.01,0.12344319820404052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,8,power_law_1.01,0.15003520250320435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,8,power_law_1.01,0.1684928059577942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,8,power_law_1.01,0.2257920026779175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,8,power_law_1.01,0.3050944089889526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,8,power_law_1.01,0.4397759914398193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,8,power_law_1.01,0.5618879795074463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,8,power_law_1.01,0.7725823879241943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,8,power_law_1.01,1.0484607696533204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,8,power_law_1.01,1.4940671920776367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,8,power_law_1.01,1.829190444946289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,8,power_law_1.01,2.498086357116699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,8,power_law_1.01,3.585036849975586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,8,power_law_1.01,7.857625579833984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,8,power_law_1.01,0.0646336019039154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,8,power_law_1.01,0.05428479909896851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,8,power_law_1.01,0.06540799736976624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,8,power_law_1.01,0.05631359815597534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,8,power_law_1.01,0.06113920211791992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,8,power_law_1.01,0.054995197057724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,8,power_law_1.01,0.0635968029499054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,8,power_law_1.01,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,8,power_law_1.01,0.06653439998626709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,8,power_law_1.01,0.05648639798164368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,8,power_law_1.01,0.06216959953308106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,8,power_law_1.01,0.054995197057724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,8,power_law_1.01,0.06074879765510559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,8,power_law_1.01,0.05422719717025757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,8,power_law_1.01,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,8,power_law_1.01,0.054176002740859985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,8,power_law_1.01,0.05822719931602478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,8,power_law_1.01,0.051923197507858274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,8,power_law_1.01,0.054150402545928955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,8,power_law_1.01,0.051692801713943484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,8,power_law_1.01,0.05319679975509643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,8,power_law_1.01,0.050335997343063356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,8,power_law_1.01,0.05862399935722351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,8,power_law_1.01,0.05682560205459595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,8,power_law_1.01,0.06204800009727478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,8,power_law_1.01,0.06316159963607788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,8,power_law_1.01,0.060038399696350095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,8,power_law_1.01,0.0580672025680542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,8,power_law_1.01,0.07912319898605347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,8,power_law_1.01,0.07364479899406433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,8,power_law_1.01,0.07883520126342773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,8,power_law_1.01,0.07124480009078979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,8,power_law_1.01,0.09105920195579528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,8,power_law_1.01,0.07617920041084289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,8,power_law_1.01,0.10963200330734253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,8,power_law_1.01,0.08832640051841736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,8,power_law_1.01,0.12799359560012818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.01,0.11105920076370239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,8,power_law_1.01,0.1762943983078003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.01,0.1381119966506958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,8,power_law_1.01,0.022111999988555908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,8,power_law_1.01,0.20156800746917725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.01,0.17984639406204223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,8,power_law_1.01,0.02385919988155365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,8,power_law_1.01,0.29123198986053467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.01,0.24380800724029542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,8,power_law_1.01,0.025843200087547303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,8,power_law_1.01,0.36762878894805906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.01,0.29950718879699706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,8,power_law_1.01,0.02287999987602234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,8,power_law_1.01,0.5308224201202393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.01,0.4084735870361328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,8,power_law_1.01,0.024217599630355836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,8,power_law_1.01,0.6734208106994629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.01,0.5381760120391845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,8,power_law_1.01,0.02512640058994293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,8,power_law_1.01,0.04811519980430603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,8,power_law_1.01,0.9308927536010743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.01,0.75763840675354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,8,power_law_1.01,0.04860160052776337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,8,power_law_1.01,1.2431167602539062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.01,1.0728447914123536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,8,power_law_1.01,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,8,power_law_1.01,1.6374591827392577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,8,power_law_1.01,1.31976318359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,8,power_law_1.01,0.042124798893928526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,8,power_law_1.01,2.521356773376465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,8,power_law_1.01,2.03002872467041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,8,power_law_1.01,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,8,power_law_1.01,0.05263360142707825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,8,power_law_1.01,3.821209716796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,8,power_law_1.01,5.03680648803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,8,power_law_1.01,0.05050240159034729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,8,power_law_1.01,0.041043201088905336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,8,power_law_1.01,0.062041598558425906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,8,power_law_1.01,0.05308799743652344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,8,power_law_1.01,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,8,power_law_1.01,0.07359359860420227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,8,power_law_1.2,0.10113279819488526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.01,0.09006080031394958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,8,power_law_1.2,0.09687039852142335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.01,0.12091519832611083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,8,power_law_1.2,0.09704959988594056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.01,0.147052800655365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,8,power_law_1.2,0.09626240134239197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.01,0.20476160049438477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,8,power_law_1.2,0.09953280091285706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.01,0.25805439949035647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,8,power_law_1.2,0.09131519794464112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.01,0.36652801036834715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,8,power_law_1.2,0.08926720023155213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.01,0.4735424041748047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,8,power_law_1.2,0.08363519906997681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.01,0.6932479858398437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,8,power_law_1.2,0.082259202003479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.01,0.9115584373474122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,8,power_law_1.2,0.07549440264701843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,8,power_law_1.2,0.07329919934272766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,8,power_law_1.01,1.124460792541504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,8,power_law_1.2,0.08880640268325805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,8,power_law_1.01,1.7715776443481446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,8,power_law_1.2,0.09351680278778077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,8,power_law_1.01,3.5054271697998045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,8,power_law_1.2,0.07962239980697632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,8,power_law_1.2,0.10894720554351807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,8,power_law_1.2,0.11486079692840576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,8,power_law_1.2,0.11648639440536498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,8,power_law_1.2,0.15294079780578612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,8,power_law_1.2,0.18175359964370727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,8,power_law_1.2,0.23935999870300292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,8,power_law_1.2,0.31724159717559813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,8,power_law_1.2,0.42668800354003905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,8,power_law_1.2,0.5679808139801026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,8,power_law_1.2,0.8050432205200195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,8,power_law_1.2,1.0326016426086426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,8,power_law_1.2,1.5411456108093262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,8,power_law_1.2,2.006175994873047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,8,power_law_1.2,2.456812858581543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,8,power_law_1.2,3.9152767181396486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,8,power_law_1.2,7.664288330078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,8,power_law_1.2,0.05324159860610962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,8,power_law_1.2,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,8,power_law_1.2,0.059033602476119995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,8,power_law_1.2,0.054764801263809205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,8,power_law_1.2,0.05975040197372437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,8,power_law_1.2,0.05937280058860779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,8,power_law_1.2,0.061343997716903687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,8,power_law_1.2,0.06330239772796631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,8,power_law_1.2,0.06121600270271301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,8,power_law_1.2,0.059411197900772095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,8,power_law_1.2,0.05774719715118408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,8,power_law_1.2,0.05395200252532959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,8,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,8,power_law_1.2,0.05092480182647705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,8,power_law_1.2,0.06204800009727478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,8,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,8,power_law_1.2,0.0582144021987915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,8,power_law_1.2,0.0808896005153656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,8,power_law_1.2,0.07906559705734253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,8,power_law_1.2,0.09039360284805298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,8,power_law_1.2,0.11079679727554322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,8,power_law_1.2,0.12744959592819213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,8,power_law_1.2,0.17289600372314454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,8,power_law_1.2,0.21025280952453612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,8,power_law_1.2,0.022163200378417968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,8,power_law_1.2,0.28080000877380373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,8,power_law_1.2,0.02260479927062988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,8,power_law_1.2,0.35201919078826904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,8,power_law_1.2,0.022431999444961548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,8,power_law_1.2,0.47591681480407716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,8,power_law_1.2,0.02279680073261261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,8,power_law_1.2,0.6830976009368896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,8,power_law_1.2,0.022860799729824067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,8,power_law_1.2,0.966163158416748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,8,power_law_1.2,0.023871999979019166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,8,power_law_1.2,1.3570752143859863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,8,power_law_1.2,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,8,power_law_1.2,1.6744768142700195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,8,power_law_1.2,0.04904319941997528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,8,power_law_1.2,2.427302360534668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,8,power_law_1.2,0.04588159918785095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,8,power_law_1.2,0.041894400119781496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,8,power_law_1.2,5.215283203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,8,power_law_1.2,0.03865599930286408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,8,power_law_1.2,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,8,power_law_1.2,0.051583999395370485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,8,power_law_1.2,0.0403903990983963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,8,power_law_1.2,0.061382400989532473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,8,power_law_1.2,0.05278080105781555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,4,balanced,0.0684746652841568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,8,power_law_1.2,0.05942400097846985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,4,balanced,0.06411199768384297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,8,power_law_1.2,0.0726527988910675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,4,balanced,0.08090133468310039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.2,0.08984959721565247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,4,balanced,0.08943466345469157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.2,0.12222080230712891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,4,balanced,0.08029333253701527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.2,0.14754559993743896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,4,balanced,0.08669867118199666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.2,0.20547199249267578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,4,balanced,0.08570667107899983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.2,0.2575232028961182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,4,balanced,0.07590400179227193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,4,balanced,0.08347733815511067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.2,0.3656512022018433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,4,balanced,0.08433066805203755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.2,0.47463040351867675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,4,balanced,0.08413867155710857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.2,0.6935232162475586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,4,balanced,0.07738133271535237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.2,0.9103743553161621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,4,balanced,0.07965866724650066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,8,power_law_1.2,1.1279616355895996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,4,balanced,0.07633066674073537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,8,power_law_1.2,1.7726272583007812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,4,balanced,0.07632533212502797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,4,balanced,0.07524266839027405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,8,power_law_1.2,3.5157630920410154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,4,balanced,0.077674667040507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,4,balanced,0.08995200196901958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,4,balanced,0.09225599964459737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,4,balanced,0.11522133151690166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,4,balanced,0.12661866346995035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,4,balanced,0.16220800081888834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,4,balanced,0.19649600982666016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,4,balanced,0.2685653368631999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,4,balanced,0.3283093372980754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,4,balanced,0.46831464767456055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,4,balanced,0.5998506546020508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,4,balanced,0.7471840381622314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,4,balanced,1.1711413065592449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,4,balanced,2.264970620473226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,4,balanced,0.049413333336512245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,4,balanced,0.050479998191197716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,4,balanced,0.053317333261171974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,4,balanced,0.05715733269850413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,4,balanced,0.05951466659704844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,4,balanced,0.05672533313433329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,4,balanced,0.06076266864935557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,4,balanced,0.058917333682378135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,4,balanced,0.06043200194835663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,4,balanced,0.061568001906077065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,4,balanced,0.05485333502292633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,4,balanced,0.05505600074927012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,4,balanced,0.05358933409055074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,4,balanced,0.05475200215975443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,4,balanced,0.05893866717815399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,4,balanced,0.05752533177534739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,4,balanced,0.061834668119748436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,4,balanced,0.07842133442560832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,4,balanced,0.084714670976003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,4,balanced,0.10393599669138591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,8,power_law_1.2,0.055340802669525145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,4,balanced,0.12264532844225566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,8,power_law_1.2,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,8,power_law_1.2,0.055820798873901366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,4,balanced,0.15588800112406412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,4,balanced,0.04613333443800608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,8,power_law_1.2,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,4,balanced,0.04703466594219208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,8,power_law_1.2,0.05299199819564819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,4,balanced,0.19218132893244425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,4,balanced,0.05765333275000254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,8,power_law_1.2,0.054067200422286986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,4,balanced,0.054789334535598755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,8,power_law_1.2,0.05149440169334411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,4,balanced,0.053717335065205894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,4,balanced,0.2637919982274373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,8,power_law_1.2,0.0472896009683609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,4,balanced,0.05438933273156484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,8,power_law_1.2,0.051609599590301515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,4,balanced,0.057328000664711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,8,power_law_1.2,0.0643775999546051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,4,balanced,0.33294934034347534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,4,balanced,0.061887999375661217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,8,power_law_1.2,0.056966400146484374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,4,balanced,0.056757330894470215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,8,power_law_1.2,0.06988160014152527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,4,balanced,0.05526400109132131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,8,power_law_1.2,0.06908800005912781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,4,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,8,power_law_1.2,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,4,balanced,0.47597332795461017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,4,balanced,0.05579199890295664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,8,power_law_1.2,0.09673600196838379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,4,balanced,0.0584799995024999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.2,0.10641920566558838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,4,balanced,0.052426666021347046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.2,0.14230400323867798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,4,balanced,0.05751466751098633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.2,0.16927360296249389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,4,balanced,0.05509333312511444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,4,balanced,0.6159786780675253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.2,0.23330559730529785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,4,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.2,0.30443520545959474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,4,balanced,0.07010133564472198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.2,0.405075216293335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.2,0.5399936199188232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,4,balanced,0.07267199953397115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.2,0.8186880111694336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,4,balanced,0.08583466211954753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.2,1.0297151565551759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,4,balanced,0.7521706422170004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,4,balanced,0.10328533252080281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,8,power_law_1.2,1.309663963317871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,4,balanced,0.12589333454767862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,8,power_law_1.2,2.1271615982055665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,8,power_law_1.2,4.2231487274169925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,4,balanced,0.15200533469518027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,4,balanced,0.2034239967664083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,4,balanced,1.1733012994130452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,4,balanced,0.25330134232838947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,4,balanced,0.3542240063349406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,4,balanced,0.4543626705805461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,4,balanced,0.5560640096664429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,4,balanced,2.2964693705240884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,4,balanced,0.8548639615376791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,4,balanced,0.021669333179791767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,4,balanced,0.023183998962243397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,4,balanced,0.023152001202106476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,4,balanced,0.024282666544119518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,4,balanced,0.02421333392461141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,4,balanced,0.024826665719350178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,4,balanced,0.028607999285062153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,4,balanced,0.026741333305835724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,4,balanced,0.050426666935284935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,4,balanced,0.050554667909940086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,4,balanced,0.05089599887530009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,4,balanced,0.04491200049718221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,4,balanced,1.6648853619893391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,4,balanced,0.043194666504859924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,4,balanced,0.03920533259709676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,4,balanced,0.04542933404445648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,4,balanced,0.041146665811538696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,4,balanced,0.045514668027559914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,4,balanced,0.05752533177534739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,4,balanced,0.06098666787147522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,4,balanced,0.08290666838486989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,4,balanced,0.10151466727256775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,4,balanced,0.13363200426101685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,4,balanced,0.1613599956035614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,4,balanced,0.22992000977198282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,4,balanced,0.29422932863235474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,4,balanced,0.42206935087839764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,4,balanced,0.5493760108947754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,4,balanced,0.6797440052032471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,4,power_law_1.01,0.07499520182609558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,4,power_law_1.01,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,4,power_law_1.01,0.0739135980606079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,4,power_law_1.01,0.07885439991950989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,4,power_law_1.01,0.0818560004234314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,4,power_law_1.01,0.07745919823646545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,4,balanced,1.0645493666330974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,4,power_law_1.01,0.07856640219688416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,4,power_law_1.01,0.07890560030937195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,4,power_law_1.01,0.07525119781494141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,4,power_law_1.01,0.07908480167388916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,4,power_law_1.01,0.0755840003490448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,4,power_law_1.01,0.08005120158195496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,4,power_law_1.01,0.07557759881019592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,4,power_law_1.01,0.077292799949646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,4,power_law_1.01,0.09557120203971863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,4,power_law_1.01,0.09889280200004577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,4,power_law_1.01,0.10292479991912842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,4,power_law_1.01,0.13068799972534179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,4,power_law_1.01,0.13524479866027833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,4,balanced,2.0863146781921387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,4,power_law_1.01,0.1937280058860779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,4,power_law_1.01,0.21821439266204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,4,power_law_1.01,0.05052800178527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,4,power_law_1.01,0.2886719942092896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,4,power_law_1.01,0.04992640018463135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,4,power_law_1.01,0.428710412979126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,4,power_law_1.01,0.053939199447631835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,4,power_law_1.01,0.5028736114501953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,4,power_law_1.01,0.05406079888343811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,4,power_law_1.01,0.7227903842926026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,4,power_law_1.01,0.05496960282325745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,4,power_law_1.01,1.0791232109069824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,4,power_law_1.01,0.05727360248565674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,4,power_law_1.01,1.4661312103271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,4,power_law_1.01,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,4,power_law_1.01,0.05505920052528381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,4,power_law_1.01,1.4041152000427246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,4,power_law_1.01,0.05545600056648255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,4,power_law_1.01,2.9975040435791014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,4,power_law_1.01,0.05545600056648255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,4,power_law_1.01,5.331187057495117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,4,power_law_1.01,0.05381760001182556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,4,power_law_1.01,0.05772160291671753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,4,power_law_1.01,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,4,power_law_1.01,0.060127997398376466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,4,power_law_1.01,0.07428479790687562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,4,power_law_1.01,0.07715200185775757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,4,power_law_1.01,0.050988799333572386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,4,power_law_1.01,0.08792960047721862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,4,power_law_1.01,0.050323200225830075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,4,power_law_1.01,0.10124800205230713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,4,power_law_1.01,0.05308160185813904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,4,power_law_1.01,0.11352319717407226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,4,power_law_1.01,0.05472000241279602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,4,power_law_1.01,0.14343680143356324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,4,power_law_1.01,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,4,power_law_1.01,0.17742079496383667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,4,power_law_1.01,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,4,power_law_1.01,0.24708480834960939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,4,power_law_1.01,0.05570560097694397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,4,power_law_1.01,0.05441280007362366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,4,power_law_1.01,0.3334144115447998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,4,power_law_1.01,0.05438719987869263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,4,power_law_1.01,0.45998082160949705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,4,power_law_1.01,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,4,power_law_1.01,0.5902463912963867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,4,power_law_1.01,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,4,power_law_1.01,0.9001152038574218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,4,power_law_1.01,0.057107198238372806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,4,power_law_1.01,0.9805760383605957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,4,power_law_1.01,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,4,power_law_1.01,1.4654080390930175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,4,power_law_1.01,0.0606656014919281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,4,power_law_1.01,2.120761680603027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,4,power_law_1.01,0.0726527988910675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,4,power_law_1.01,0.06839039921760559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,4,power_law_1.01,4.826764678955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,4,power_law_1.01,0.07725440263748169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,4,power_law_1.01,0.08996480107307434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.01,0.10791679620742797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.01,0.12753280401229858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.01,0.1427199959754944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.01,0.20500481128692627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.01,0.24567680358886718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.01,0.3563136100769043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.01,0.46202239990234373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.01,0.6088511943817139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.01,1.0816960334777832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,4,power_law_1.01,1.1430208206176757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,4,power_law_1.01,1.89050235748291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,4,power_law_1.01,3.192678451538086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,4,power_law_1.01,0.020608000457286835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,4,power_law_1.01,0.020684799551963805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,4,power_law_1.01,0.020960000157356263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,4,power_law_1.01,0.021324799954891206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,4,power_law_1.01,0.021299199759960176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,4,power_law_1.01,0.02237440049648285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,4,power_law_1.01,0.026624000072479247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,4,power_law_1.01,0.025817599892616273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,4,power_law_1.2,0.06926079988479614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,4,power_law_1.01,0.04952319860458374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,4,power_law_1.2,0.07494400143623352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,4,power_law_1.01,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,4,power_law_1.2,0.0769919991493225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,4,power_law_1.2,0.07410560250282287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,4,power_law_1.01,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,4,power_law_1.2,0.07700480222702026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,4,power_law_1.01,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,4,power_law_1.2,0.07651839852333069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,4,power_law_1.01,0.0419840008020401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,4,power_law_1.2,0.07849599719047547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,4,power_law_1.01,0.038867199420928956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,4,power_law_1.2,0.07838720083236694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,4,power_law_1.2,0.08005120158195496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,4,power_law_1.01,0.04431360065937042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,4,power_law_1.2,0.0754688024520874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,4,power_law_1.01,0.03997440040111542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,4,power_law_1.2,0.07530879974365234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,4,power_law_1.01,0.044531199336051944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,4,power_law_1.2,0.07495679855346679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,4,power_law_1.01,0.05671039819717407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,4,power_law_1.2,0.07819520235061646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.01,0.06133120059967041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,4,power_law_1.2,0.08117120265960694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.01,0.08168320059776306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,4,power_law_1.2,0.09416319727897644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.01,0.10047359466552734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,4,power_law_1.2,0.0906175971031189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.01,0.1345023989677429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,4,power_law_1.2,0.11340160369873047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.01,0.16069120168685913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,4,power_law_1.2,0.13344000577926635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,4,power_law_1.2,0.151692795753479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.01,0.23009281158447265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,4,power_law_1.2,0.1611840009689331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.01,0.2925695896148682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,4,power_law_1.2,0.22218880653381348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.01,0.4211455821990967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,4,power_law_1.2,0.0546176016330719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,4,power_law_1.2,0.2431488037109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,4,power_law_1.2,0.04978559911251068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.01,0.5502399921417236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,4,power_law_1.2,0.4413504123687744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,4,power_law_1.2,0.053324800729751584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,4,power_law_1.01,0.6818496227264405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,4,power_law_1.2,0.6389056205749511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,4,power_law_1.2,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,4,power_law_1.01,1.067404842376709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,4,power_law_1.2,0.8573760032653809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,4,power_law_1.2,0.05768960118293762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,4,power_law_1.01,2.091129684448242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,4,power_law_1.2,1.2097408294677734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,4,power_law_1.2,0.05571200251579285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,4,power_law_1.2,0.05631359815597534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,4,power_law_1.2,1.439680004119873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,4,power_law_1.2,0.0545087993144989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,4,power_law_1.2,1.9488447189331055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,4,power_law_1.2,0.059648001194000246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,4,power_law_1.2,3.254207992553711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,4,power_law_1.2,0.05416319966316223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,4,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,4,power_law_1.2,7.6226753234863285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,4,power_law_1.2,0.057792001962661745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,4,power_law_1.2,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,4,power_law_1.2,0.06369280219078063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,4,power_law_1.2,0.07354239821434021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,4,power_law_1.2,0.0789247989654541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,4,power_law_1.2,0.08990079760551453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,4,power_law_1.2,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,4,power_law_1.2,0.09957119822502136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,4,power_law_1.2,0.05534719824790955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,4,power_law_1.2,0.04749439954757691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,4,power_law_1.2,0.11871999502182007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,4,power_law_1.2,0.05422080159187317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,4,power_law_1.2,0.15544320344924928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,4,power_law_1.2,0.05424000024795532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,4,power_law_1.2,0.17480319738388062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,4,power_law_1.2,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,4,power_law_1.2,0.27984640598297117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,4,power_law_1.2,0.055257600545883176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,4,power_law_1.2,0.30674560070037843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,4,power_law_1.2,0.055212801694869994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,4,power_law_1.2,0.4521279811859131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,4,power_law_1.2,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,4,power_law_1.2,0.6287487983703614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,4,power_law_1.2,0.05451520085334778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,4,power_law_1.2,0.8375359535217285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,4,power_law_1.2,0.05381119847297668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,4,power_law_1.2,1.3379136085510255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,4,power_law_1.2,0.059961599111557004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,4,power_law_1.2,1.641983985900879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,4,power_law_1.2,0.05851519703865051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,4,power_law_1.2,2.5664064407348635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,4,power_law_1.2,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,4,power_law_1.2,0.07128319740295411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,4,power_law_1.2,5.052479934692383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,4,power_law_1.2,0.07104640007019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,4,power_law_1.2,0.08147199749946595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,4,power_law_1.2,0.09567999839782715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.2,0.11063679456710815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.2,0.11934720277786255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.2,0.15615999698638916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.2,0.24018559455871583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.2,0.2964544057846069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.2,0.41544318199157715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.2,0.5206719875335694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.2,0.7612224102020264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.2,1.0979455947875976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,4,power_law_1.2,0.980684757232666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,4,power_law_1.2,1.9957311630249024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,4,power_law_1.2,3.3898815155029296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,4,power_law_1.2,0.019840000569820403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,2,balanced,0.05095999936262766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,2,balanced,0.05931733548641205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,4,power_law_1.2,0.021267199516296388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,2,balanced,0.07223999996980031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,4,power_law_1.2,0.021612800657749176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,2,balanced,0.08355733752250671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,4,power_law_1.2,0.022169600427150726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,2,balanced,0.07442666590213776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,4,power_law_1.2,0.022009600698947907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,2,balanced,0.07826133569081624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,4,power_law_1.2,0.022284799814224245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,2,balanced,0.07859733204046886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,4,power_law_1.2,0.026131200790405273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,2,balanced,0.07386666536331177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,4,power_law_1.2,0.02595199942588806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,2,balanced,0.07891199986139934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,4,power_law_1.2,0.049958398938179015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,2,balanced,0.07643733421961467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,4,power_law_1.2,0.05074560046195984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,2,balanced,0.07934399942557017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,4,power_law_1.2,0.050374400615692136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,2,balanced,0.07825066645940144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,4,power_law_1.2,0.042208001017570496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,2,balanced,0.07154133419195811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,4,power_law_1.2,0.04156799912452698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,2,balanced,0.07234666744867961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,4,power_law_1.2,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,2,balanced,0.07725866635640462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,4,power_law_1.2,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,2,balanced,0.07635200023651123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,4,power_law_1.2,0.039699199795722964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,2,balanced,0.07950933277606964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,4,power_law_1.2,0.04359039962291718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,2,balanced,0.09225066502888997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,4,power_law_1.2,0.0561024010181427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,2,balanced,0.09388800462086995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.2,0.061408001184463504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,2,balanced,0.1229759951432546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.2,0.08161280155181885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.2,0.09957119822502136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,2,balanced,0.13481066624323526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,2,balanced,0.04215466479460398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.2,0.13457920551300048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,2,balanced,0.0459146648645401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,2,balanced,0.17341333627700806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.2,0.16024960279464723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,2,balanced,0.05300266544024149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.2,0.23138558864593506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,2,balanced,0.212826669216156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,2,balanced,0.0506879985332489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.2,0.29219200611114504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,2,balanced,0.05738666653633118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.2,0.42081279754638673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,2,balanced,0.05397333204746246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,2,balanced,0.29064534107844037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.2,0.5509696006774902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,2,balanced,0.053258667389551796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,2,balanced,0.05323733389377594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,4,power_law_1.2,0.6796160221099854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,2,balanced,0.05499733487764994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,2,balanced,0.35767467816670734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,4,power_law_1.2,1.0623040199279785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,2,balanced,0.05517866710821787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,4,power_law_1.2,2.08666877746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,2,balanced,0.05175999800364176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,2,balanced,0.052186667919158936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,2,balanced,0.5226613283157349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,2,balanced,0.057002668579419456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,2,balanced,0.05499200026194254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,2,balanced,0.05936533212661743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,2,balanced,0.06166933476924896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,2,balanced,0.6706293423970541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,2,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,2,balanced,0.06698666512966156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,2,balanced,0.05124266445636749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,2,balanced,0.0830026666323344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,2,balanced,0.05829333265622457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,2,balanced,0.08943466345469157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,2,balanced,0.057328000664711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,2,balanced,0.05367999772230784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,2,balanced,0.11662399768829346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,2,balanced,0.0554613322019577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,2,balanced,0.8244586785634359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,2,balanced,0.13237866759300232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,2,balanced,0.05704000095526377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,2,balanced,0.056832000613212585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,2,balanced,0.17301867405573526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,2,balanced,0.05727999905745188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,2,balanced,0.05492266515890757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,2,balanced,0.21313599745432535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,2,balanced,0.053717335065205894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,2,balanced,0.050698667764663696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,2,balanced,0.30081599950790405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,2,balanced,0.0552106648683548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,2,balanced,0.05471999943256378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,2,balanced,1.2956480185190837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,2,balanced,0.05508266886075338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,2,balanced,0.379472017288208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,2,balanced,0.059487998485565186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,2,balanced,0.059674665331840515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,2,balanced,0.07473066449165344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,2,balanced,0.5516266822814941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,2,balanced,0.07743466893831889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,2,balanced,0.09450667103131612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,2,balanced,0.11099732915560405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,2,balanced,0.7159413496653239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,2,balanced,0.14337600270907083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,2,balanced,0.1720906694730123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,2,balanced,0.8770186901092529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,2,balanced,0.2398293415705363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,2,balanced,2.5687840779622397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,2,balanced,0.29803200562795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,2,balanced,1.3698453903198242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,2,balanced,0.42100266615549725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,2,balanced,0.5477919975916544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,2,balanced,0.6702133019765218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,2,balanced,2.6825599670410156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,2,balanced,1.04966402053833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,2,balanced,2.052549362182617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,2,balanced,0.02162133405605952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,2,balanced,0.022863999009132385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,2,balanced,0.022650666534900665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,2,balanced,0.022618666291236877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,2,balanced,0.02388266722361247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,2,balanced,0.022597332795461018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,2,balanced,0.023733332753181458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,2,balanced,0.024671999116738636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,2,balanced,0.026250667870044708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,2,balanced,0.026314665873845417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,2,balanced,0.027808000644048054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,2,balanced,0.050570666790008545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,2,balanced,0.050810664892196655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,2,balanced,0.05309333403905233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,2,balanced,0.04032533367474874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,2,balanced,0.03866666555404663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,2,balanced,0.04277333120505015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,2,balanced,0.04696000119050344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,2,balanced,0.04985600213209788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,2,balanced,0.06698666512966156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,2,balanced,0.07470933099587758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,2,balanced,0.09974933664004008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,2,balanced,0.12193600336710612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,2,balanced,0.16800000270207724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,2,balanced,0.2037066618601481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,2,power_law_1.01,0.05406079888343811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,2,power_law_1.01,0.057017600536346434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,2,power_law_1.01,0.06238719820976257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,2,balanced,0.2979733347892761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,2,power_law_1.01,0.0735360026359558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,2,power_law_1.01,0.07639679908752442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,2,power_law_1.01,0.07444480061531067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,2,power_law_1.01,0.0760703980922699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,2,balanced,0.3810133139292399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,2,power_law_1.01,0.07485439777374267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,2,power_law_1.01,0.07485439777374267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,2,power_law_1.01,0.0743552029132843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,2,power_law_1.01,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,2,power_law_1.01,0.07567359805107117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,2,balanced,0.46421865622202557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,2,power_law_1.01,0.07656959891319275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,2,power_law_1.01,0.07582719922065735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,2,power_law_1.01,0.08298879861831665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,2,power_law_1.01,0.08413439989089966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,2,power_law_1.01,0.08828799724578858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,2,power_law_1.01,0.10758399963378906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,2,balanced,0.7243306636810303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,2,power_law_1.01,0.1249343991279602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,2,power_law_1.01,0.15611519813537597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,2,power_law_1.01,0.19253120422363282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,2,power_law_1.01,0.24531838893890381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,2,power_law_1.01,0.266374397277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,2,power_law_1.01,0.3614016056060791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,2,power_law_1.01,0.5136127948760987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,2,power_law_1.01,0.04302720129489899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,2,power_law_1.01,0.794707202911377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,2,power_law_1.01,0.04901759922504425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,2,power_law_1.01,1.1387455940246582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,2,power_law_1.01,0.048102399706840514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,2,power_law_1.01,0.05351679921150208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,2,power_law_1.01,1.25349760055542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,2,power_law_1.01,0.05371519923210144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,2,balanced,1.415514628092448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,2,power_law_1.01,2.1465087890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,2,power_law_1.01,0.05429760217666626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,2,power_law_1.01,3.568288040161133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,2,power_law_1.01,0.05482239723205566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,2,power_law_1.01,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,2,power_law_1.01,0.05553280115127564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,2,power_law_1.01,0.055193597078323366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,2,power_law_1.01,0.054713600873947145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,2,power_law_1.01,0.05504639744758606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,2,power_law_1.01,0.06116480231285095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,2,power_law_1.01,0.059487998485565186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,2,power_law_1.01,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,2,power_law_1.01,0.06746240258216858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,2,power_law_1.01,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,2,power_law_1.01,0.07431039810180665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,2,power_law_1.01,0.0492031991481781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,2,power_law_1.01,0.07607679963111877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,2,power_law_1.01,0.05281919836997986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,2,power_law_1.01,0.09110400080680847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,2,power_law_1.01,0.10828800201416015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,2,power_law_1.01,0.055219197273254396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,2,power_law_1.01,0.14051200151443483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,2,power_law_1.01,0.1646463990211487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,2,power_law_1.01,0.21653120517730712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,2,power_law_1.01,0.26371839046478274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,2,power_law_1.01,0.3898688077926636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,2,power_law_1.01,0.4947648048400879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,2,power_law_1.01,0.7105792045593262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,2,power_law_1.01,0.9253631591796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,2,power_law_1.01,1.1736960411071777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,2,power_law_1.01,1.8247615814208984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,2,power_law_1.01,3.619014358520508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,2,power_law_1.01,0.019270400702953338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,2,power_law_1.01,0.02110079973936081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,2,power_law_1.01,0.020627200603485107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,2,power_law_1.2,0.05428479909896851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,2,power_law_1.01,0.02091519981622696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,2,power_law_1.2,0.05960320234298706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,2,power_law_1.01,0.021792000532150267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,2,power_law_1.2,0.05976319909095764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,2,power_law_1.01,0.02003840059041977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,2,power_law_1.2,0.07490559816360473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,2,power_law_1.2,0.07850239872932434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,2,power_law_1.01,0.021267199516296388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,2,power_law_1.2,0.07287039756774902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,2,power_law_1.01,0.022860799729824067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,2,power_law_1.2,0.07379840016365051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,2,power_law_1.01,0.024876800179481507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,2,power_law_1.2,0.0745024025440216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,2,power_law_1.01,0.023955200612545014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,2,power_law_1.2,0.0755840003490448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,2,power_law_1.01,0.02752000093460083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,2,power_law_1.2,0.07409279942512512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,2,power_law_1.01,0.04800640046596527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,2,power_law_1.2,0.0715071976184845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,2,power_law_1.01,0.04922879934310913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,2,power_law_1.2,0.07264000177383423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,2,power_law_1.01,0.05098239779472351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,2,power_law_1.2,0.0756608009338379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,2,power_law_1.01,0.03985919952392578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,2,power_law_1.2,0.0787775993347168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,2,power_law_1.01,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,2,power_law_1.2,0.08511360287666321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,2,power_law_1.01,0.040755200386047366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,2,power_law_1.2,0.08254719972610473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,2,power_law_1.01,0.04425599873065948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,2,power_law_1.2,0.09011840224266052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.01,0.047116801142692566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,2,power_law_1.2,0.11316479444503784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.01,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,2,power_law_1.2,0.127839994430542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.01,0.07368320226669312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,2,power_law_1.2,0.15482239723205565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.01,0.09816319942474365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,2,power_law_1.2,0.19290879964828492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.01,0.1218559980392456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,2,power_law_1.2,0.23497600555419923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.01,0.1674880027770996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,2,power_law_1.2,0.307423996925354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.01,0.20453760623931885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,2,power_law_1.2,0.045414400100708005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,2,power_law_1.2,0.4453887939453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.01,0.2975167989730835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,2,power_law_1.2,0.04423680007457733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,2,power_law_1.2,0.49813117980957033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.01,0.38109440803527833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,2,power_law_1.2,0.046675199270248414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,2,power_law_1.2,0.8131072044372558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,2,power_law_1.01,0.46428799629211426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,2,power_law_1.2,0.05288959741592407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,2,power_law_1.2,0.9124159812927246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,2,power_law_1.01,0.7250112056732178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,2,power_law_1.2,0.05602560043334961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,2,power_law_1.2,1.3306048393249512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,2,power_law_1.01,1.416774368286133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,2,power_law_1.2,0.053504002094268796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,2,power_law_1.2,2.297920036315918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,2,power_law_1.2,0.05446400046348572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,2,power_law_1.2,3.641875076293945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,2,power_law_1.2,0.0546176016330719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,2,power_law_1.2,0.05549439787864685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,2,power_law_1.2,0.05560960173606873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,2,power_law_1.2,0.055308800935745236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,2,power_law_1.2,0.058956801891326904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,2,power_law_1.2,0.060224002599716185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,2,power_law_1.2,0.06047359704971313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,2,power_law_1.2,0.06940799951553345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,2,power_law_1.2,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,2,power_law_1.2,0.07763839960098266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,2,power_law_1.2,0.09707520008087159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,2,power_law_1.2,0.10953600406646728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,2,power_law_1.2,0.1322175979614258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,2,power_law_1.2,0.1687872052192688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,2,power_law_1.2,0.2227776050567627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,2,power_law_1.2,0.28636798858642576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,2,power_law_1.2,0.3797247886657715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,2,power_law_1.2,0.5135551929473877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,2,power_law_1.2,0.6861055850982666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,2,power_law_1.2,0.9358464241027832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,2,power_law_1.2,1.1980607986450196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,2,power_law_1.2,1.9695680618286133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,2,power_law_1.2,3.7501888275146484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,2,power_law_1.01,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,2,power_law_1.01,0.05621119737625122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,2,power_law_1.01,0.05623040199279785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,2,power_law_1.01,0.0542464017868042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,2,power_law_1.01,0.05367040038108826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,2,power_law_1.01,0.05518720149993896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,2,power_law_1.01,0.05551999807357788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,2,power_law_1.01,0.06082559823989868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,2,power_law_1.01,0.062035202980041504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,2,power_law_1.01,0.06569600105285645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,2,power_law_1.01,0.06875519752502442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,2,power_law_1.01,0.07483519911766053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,2,power_law_1.01,0.08215680122375488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.01,0.10190080404281616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.01,0.1210368037223816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.01,0.12870399951934813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.01,0.1803007960319519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.01,0.2570496082305908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.01,0.26311678886413575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.01,0.41093759536743163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.01,0.5207680225372314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.01,0.793555212020874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,2,power_law_1.01,0.9771264076232911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,2,power_law_1.01,1.3132287979125976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,2,power_law_1.01,2.6198911666870117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,2,power_law_1.2,0.019392000138759614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,2,power_law_1.2,0.04940159916877747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,2,power_law_1.2,0.021011200547218323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,2,power_law_1.2,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,2,power_law_1.2,0.02144639939069748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,2,power_law_1.2,0.0473471999168396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,2,power_law_1.2,0.021452799439430237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,2,power_law_1.2,0.05548160076141358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,2,power_law_1.2,0.022118400037288665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,2,power_law_1.2,0.0548799991607666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,2,power_law_1.2,0.021932800114154816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,2,power_law_1.2,0.05609599947929382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,2,power_law_1.2,0.021753600239753722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,2,power_law_1.2,0.05614079833030701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,2,power_law_1.2,0.02271360009908676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,2,power_law_1.2,0.05560960173606873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,2,power_law_1.2,0.05602560043334961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,2,power_law_1.2,0.02412160038948059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,2,power_law_1.2,0.05459200143814087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,2,power_law_1.2,0.024307200312614442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,2,power_law_1.2,0.05416960120201111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,2,power_law_1.2,0.02584959864616394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,2,power_law_1.2,0.05749760270118713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,2,power_law_1.2,0.06060799956321716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,2,power_law_1.2,0.06021760106086731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,2,power_law_1.2,0.06744959950447083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,2,power_law_1.2,0.06781439781188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,2,power_law_1.2,0.07607679963111877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,2,power_law_1.2,0.08377599716186523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.2,0.10126719474792481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.2,0.11682560443878173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.2,0.13932160139083863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.2,0.18926080465316772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.2,0.2494976043701172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.2,0.31767680644989016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,1,balanced,0.048357332746187844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.2,0.4035071849822998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,1,balanced,0.056602666775385536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.2,0.5868351936340332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,1,balanced,0.07223466535409291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,1,balanced,0.07751999795436859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.2,0.7931583881378174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,1,balanced,0.07252266506354015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,2,power_law_1.2,1.03372802734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,1,balanced,0.07365333537260692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,2,power_law_1.2,1.6343040466308594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,1,balanced,0.07231466472148895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,2,power_law_1.2,3.1745920181274414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,1,balanced,0.0706826647122701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,1,balanced,0.07561600208282471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,1,balanced,0.07549866537253062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,1,balanced,0.07386666536331177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,1,balanced,0.0765173335870107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,1,balanced,0.07469333211580913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,1,balanced,0.07584000130494435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,1,balanced,0.08352532982826233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,1,balanced,0.082096000512441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,1,balanced,0.08642666538556416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,1,balanced,0.1000426709651947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,1,balanced,0.10585066676139832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,1,balanced,0.13965866963068643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,1,balanced,0.15613866845766702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,1,balanced,0.21011199553807577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,1,balanced,0.2587626576423645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,1,balanced,0.36213866869608563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,1,balanced,0.44979198773701984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,1,balanced,0.6548266808191935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,1,balanced,0.03999999910593033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,1,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,1,balanced,0.05142400165398916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,1,balanced,0.05550933380921682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,1,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,1,balanced,0.05377600093682607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,1,balanced,0.8404320081075033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,1,balanced,0.05737066765626272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,1,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,1,balanced,0.0533493310213089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,1,balanced,0.05348266661167145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,1,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,1,balanced,1.0310719807942708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,1,balanced,0.05653333167235056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,1,balanced,0.05927466849486033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,1,balanced,0.059978668888409935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,1,balanced,0.06779199838638306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,1,balanced,0.06922133266925812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,1,balanced,0.07541333138942719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,1,balanced,0.09551999966303508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,1,balanced,1.613375981648763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,1,balanced,0.10609066486358643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,1,balanced,0.1363146702448527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,1,balanced,0.15804266929626465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,1,balanced,0.21275200446446738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,1,balanced,0.27049599091211957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,1,balanced,0.3827040195465088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,1,balanced,0.4821653366088867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,1,balanced,3.2180372873942056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,1,balanced,0.7028853098551432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,1,balanced,0.9147893587748209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,1,balanced,1.1300373077392578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,1,balanced,0.04407466451327006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,1,balanced,1.7665600776672363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,1,balanced,0.05113600194454193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,1,balanced,0.05690666536490122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,1,balanced,0.05919999877611796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,1,balanced,0.05518400172392527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,1,balanced,0.06010666489601135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,1,balanced,0.05938133100668589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,1,balanced,0.062208001812299095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,1,balanced,0.05518400172392527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,1,balanced,0.05686399837334951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,1,balanced,0.06125866870085398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,1,balanced,0.05932799975077311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,1,balanced,0.05949333310127258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,1,balanced,0.061162665486335754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,1,balanced,3.482799847920736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,1,balanced,0.06589866677920024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,1,balanced,0.06610133250554402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,1,balanced,0.06820799907048543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,1,balanced,0.09066667159398396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,1,balanced,0.09689066807428996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,1,balanced,0.11909332871437073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,1,balanced,0.14100266496340433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,1,balanced,0.18610133727391562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,2,power_law_1.2,0.04871039986610413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,1,balanced,0.23164266347885132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,2,power_law_1.2,0.04894079864025116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,2,power_law_1.2,0.04984320104122162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,2,power_law_1.2,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,1,balanced,0.3232319951057434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,2,power_law_1.2,0.03770880103111267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,2,power_law_1.2,0.04000000059604645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,2,power_law_1.2,0.044198399782180785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,1,balanced,0.4161813259124756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.2,0.04803839921951294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.2,0.06720640063285828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.2,0.07308160066604615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.2,0.0987775981426239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.2,0.12186239957809449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,1,balanced,0.5916586716969808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.2,0.16817920207977294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.2,0.20467839241027833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.2,0.29448320865631106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.2,0.3814016103744507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,2,power_law_1.2,0.4641280174255371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,1,balanced,0.777184009552002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,2,power_law_1.2,0.7236480236053466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,2,power_law_1.2,1.4167488098144532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,1,balanced,0.9597280025482178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,1,balanced,0.0216799999276797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,1,balanced,0.023567999402681988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,1,balanced,0.022218666970729828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,1,balanced,0.022197333474953968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,1,balanced,0.02274133265018463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,1,balanced,1.4979519844055176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,1,balanced,0.02460266649723053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,1,balanced,0.02388266722361247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,1,balanced,0.02459733436505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,1,balanced,0.024447999894618988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,1,balanced,0.024527999262015026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,1,balanced,0.02587733417749405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,1,balanced,0.026501332720120747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,1,balanced,0.02834133307139079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,1,balanced,0.028783999383449554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,1,balanced,0.052709331115086876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,1,balanced,0.05263466636339823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,1,balanced,0.05494933327039083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,1,balanced,0.044938668608665466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,1,balanced,0.04865066707134247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,1,balanced,0.054832001527150474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,1,balanced,0.06402133405208588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,1,balanced,0.08821866909662883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,1,balanced,2.944250742594401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,1,balanced,0.1014453371365865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,1,balanced,0.13531733552614847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,1,balanced,0.16894932587941489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,1,power_law_1.01,0.05066879987716675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,1,power_law_1.01,0.05272960066795349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,1,power_law_1.01,0.05464320182800293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,1,balanced,0.24154132604599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,1,power_law_1.01,0.06858239769935608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,1,power_law_1.01,0.07127040028572082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,1,power_law_1.01,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,1,power_law_1.01,0.07178879976272583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,1,power_law_1.01,0.07085440158843995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,1,balanced,0.29946666955947876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,1,power_law_1.01,0.07388160228729249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,1,power_law_1.01,0.07118080258369446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,1,power_law_1.01,0.07233920097351074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,1,power_law_1.01,0.07177600264549255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,1,power_law_1.01,0.07436800003051758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,1,balanced,0.3678719997406006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,1,power_law_1.01,0.07657600045204163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,1,power_law_1.01,0.08250880241394043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,1,power_law_1.01,0.08526719808578491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,1,power_law_1.01,0.08844159841537476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,1,power_law_1.01,0.1016767978668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,1,power_law_1.01,0.1241152048110962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,1,power_law_1.01,0.1411520004272461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,1,balanced,0.5703999996185303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,1,power_law_1.01,0.16253440380096434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,1,power_law_1.01,0.21364479064941405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,1,power_law_1.01,0.2638144016265869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,1,power_law_1.01,0.37223680019378663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,1,power_law_1.01,0.4546688079833984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,1,power_law_1.01,0.6577216148376465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,1,power_law_1.01,0.8476415634155273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,1,power_law_1.01,1.0516415596008302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,1,power_law_1.01,1.6356096267700195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,1,power_law_1.01,3.2336448669433593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,1,balanced,1.0944960117340088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,1,power_law_1.01,0.04188160002231598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,1,power_law_1.01,0.04296320080757141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,1,power_law_1.01,0.04583039879798889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,1,power_law_1.01,0.05299199819564819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,1,power_law_1.01,0.05236480236053467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,1,power_law_1.01,0.05616000294685364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,1,power_law_1.01,0.05588480234146118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,1,power_law_1.01,0.05538560152053833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,1,power_law_1.01,0.054611200094223024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,1,power_law_1.01,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,1,power_law_1.01,0.05549439787864685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,1,power_law_1.01,0.05594239830970764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,1,power_law_1.01,0.058387202024459836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,1,power_law_1.01,0.060761600732803345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,1,power_law_1.01,0.06906239986419678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,1,power_law_1.01,0.07141759991645813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,1,power_law_1.01,0.07862399816513062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,1,power_law_1.01,0.09560959935188293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,1,power_law_1.01,0.11282559633255004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,1,power_law_1.01,0.1371392011642456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,1,power_law_1.01,0.15794559717178344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,1,power_law_1.01,0.21482880115509034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,1,power_law_1.01,0.26935040950775146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,1,power_law_1.01,0.3838912010192871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,1,power_law_1.01,0.4884479999542236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,1,power_law_1.01,0.7034751892089843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,1,power_law_1.01,0.9185088157653809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,1,power_law_1.01,1.1343680381774903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,1,power_law_1.01,1.7711551666259766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,1,power_law_1.01,3.483500671386719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,1,power_law_1.01,0.041433599591255185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,1,power_law_1.01,0.04674560129642487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,1,power_law_1.01,0.046937599778175354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,1,power_law_1.01,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,1,power_law_1.01,0.05809919834136963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,1,power_law_1.01,0.058355200290679934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,1,power_law_1.01,0.06007680296897888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,1,power_law_1.01,0.05800960063934326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,1,power_law_1.01,0.05813760161399841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,1,power_law_1.01,0.0573311984539032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,1,power_law_1.01,0.05793280005455017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,1,power_law_1.01,0.05858560204505921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,1,power_law_1.01,0.06110079884529114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,1,power_law_1.01,0.06268799901008607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,1,power_law_1.01,0.06915199756622314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,1,power_law_1.01,0.0707264006137848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,1,power_law_1.01,0.07936000227928161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,1,power_law_1.01,0.0919871985912323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.01,0.10213119983673095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.01,0.12332799434661865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.01,0.1451840043067932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.01,0.19442559480667115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.01,0.24043519496917726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.01,0.32803199291229246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.01,0.4161407947540283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.01,0.5945919990539551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.01,0.7815680027008056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.01,0.9597760200500488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,1,power_law_1.01,1.4913408279418945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,1,power_law_1.01,2.928326416015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,1,power_law_1.2,0.050470399856567386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,1,power_law_1.01,0.01966080069541931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,1,power_law_1.2,0.05253120064735413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,1,power_law_1.01,0.02040960043668747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,1,power_law_1.2,0.05635200142860412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,1,power_law_1.01,0.021273599565029146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,1,power_law_1.2,0.07004160284996033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,1,power_law_1.01,0.020851199328899384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,1,power_law_1.2,0.07246720194816589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,1,power_law_1.01,0.021580800414085388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,1,power_law_1.2,0.07326719760894776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,1,power_law_1.01,0.02250880002975464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,1,power_law_1.2,0.07015680074691773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,1,power_law_1.01,0.022623999416828154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,1,power_law_1.01,0.022547200322151184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,1,power_law_1.01,0.02215680032968521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,1,power_law_1.01,0.023155200481414794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,1,power_law_1.01,0.024223999679088594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,1,power_law_1.01,0.025772801041603087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,1,power_law_1.01,0.026528000831604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,1,power_law_1.01,0.027993598580360414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,1,power_law_1.01,0.05135999917984009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,1,power_law_1.01,0.05262079834938049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,1,power_law_1.01,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,1,power_law_1.01,0.04345600008964538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.01,0.04717440009117126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.01,0.05490559935569763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.01,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.01,0.08642560243606567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.01,0.09941120147705078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.01,0.13389439582824708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.01,0.16925439834594727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.01,0.23783040046691895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.01,0.29833600521087644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.01,0.36871678829193116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,1,power_law_1.01,0.5679935932159423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,1,power_law_1.01,1.0947839736938476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,1,power_law_1.2,0.0406143993139267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,1,power_law_1.2,0.04361599981784821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,1,power_law_1.2,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,1,power_law_1.2,0.05329279899597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,1,power_law_1.2,0.05392000079154968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,1,power_law_1.2,0.05400320291519165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,1,power_law_1.2,0.05497599840164184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,1,power_law_1.2,0.05560960173606873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,1,power_law_1.2,0.05441280007362366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,1,power_law_1.2,0.054790401458740236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,1,power_law_1.2,0.05692160129547119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,1,power_law_1.2,0.060166400671005246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,1,power_law_1.2,0.05937920212745666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,1,power_law_1.2,0.062457597255706786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,1,power_law_1.2,0.06901760101318359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,1,power_law_1.2,0.07190399765968322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,1,power_law_1.2,0.07973759770393371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,1,power_law_1.2,0.09567999839782715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,1,power_law_1.2,0.11174399852752685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,1,power_law_1.2,0.1374848008155823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,1,power_law_1.2,0.1587455987930298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,1,power_law_1.2,0.21587839126586914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,1,power_law_1.2,0.27112319469451907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,1,power_law_1.2,0.38426880836486815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,1,power_law_1.2,0.4849599838256836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,1,power_law_1.2,0.7040703773498536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,1,power_law_1.2,0.916262435913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,1,power_law_1.2,1.1357312202453613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,1,power_law_1.2,1.772652816772461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,1,power_law_1.2,3.47523193359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,1,power_law_1.2,0.04192639887332916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,1,power_law_1.2,0.045126399397850035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,1,power_law_1.2,0.047193598747253415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,1,power_law_1.2,0.058982402086257935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,1,power_law_1.2,0.05671039819717407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,1,power_law_1.2,0.05610880255699158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,1,power_law_1.2,0.05671679973602295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,1,power_law_1.2,0.06062080264091492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,1,power_law_1.2,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,1,power_law_1.2,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,1,power_law_1.2,0.058387202024459836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,1,power_law_1.2,0.05852800011634827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,1,power_law_1.2,0.06144639849662781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,1,power_law_1.2,0.06223999857902527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,1,power_law_1.2,0.06870399713516236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,1,power_law_1.2,0.07178239822387696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,1,power_law_1.2,0.07768319845199585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,1,power_law_1.2,0.09143679738044738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.2,0.10302720069885254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.2,0.125164794921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.2,0.14485759735107423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.2,0.19198720455169677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.2,0.23841280937194825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.2,0.3302720069885254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.2,0.4146240234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.2,0.5980480194091797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.2,0.7765376091003418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.2,0.9576831817626953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,1,power_law_1.2,1.489510440826416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,1,power_law_1.2,2.9204736709594727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,1,power_law_1.2,0.01913599967956543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,1,power_law_1.2,0.021414400637149812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,1,power_law_1.2,0.019731199741363524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,1,power_law_1.2,0.02022400051355362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,1,power_law_1.2,0.02136320024728775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,1,power_law_1.2,0.022700800001621245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,1,power_law_1.2,0.021030400693416596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,1,power_law_1.2,0.022598400712013245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,1,power_law_1.2,0.022720000147819518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,1,power_law_1.2,0.023071999847888946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,1,power_law_1.2,0.02375040054321289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,1,power_law_1.2,0.025894400477409363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,1,power_law_1.2,0.027212798595428467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,1,power_law_1.2,0.02865920066833496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,1,power_law_1.2,0.05130879878997803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,1,power_law_1.2,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,1,power_law_1.2,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,1,power_law_1.2,0.04311679899692535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.2,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.2,0.055155199766159055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.2,0.06312959790229797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.2,0.08579840064048767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.2,0.0988864004611969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.2,0.13337600231170654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.2,0.16785279512405396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,1,power_law_1.2,0.07064319849014282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.2,0.23809919357299805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,1,power_law_1.2,0.07074559926986694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.2,0.29832959175109863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,1,power_law_1.2,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.2,0.3699136018753052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,1,power_law_1.2,0.07264639735221863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,1,power_law_1.2,0.5700416088104248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,1,power_law_1.2,0.07279999852180481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,1,power_law_1.2,1.0940863609313964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,1,power_law_1.2,0.07312639951705932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,1,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,1,power_law_1.2,0.08230400085449219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,1,power_law_1.2,0.08273919820785522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,1,power_law_1.2,0.08965759873390197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,1,power_law_1.2,0.10023679733276367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,1,power_law_1.2,0.1256832003593445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,1,power_law_1.2,0.1446272015571594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,1,power_law_1.2,0.16387200355529785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,1,power_law_1.2,0.21151359081268312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,1,power_law_1.2,0.2640064001083374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,1,power_law_1.2,0.3678528070449829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,1,power_law_1.2,0.45261440277099607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,1,power_law_1.2,0.6593791961669921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,1,power_law_1.2,0.8454655647277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,1,power_law_1.2,1.0438400268554688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,1,power_law_1.2,1.638528060913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,1,power_law_1.2,3.224473571777344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,32,balanced,0.08342933654785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,32,balanced,0.09010666608810425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,32,balanced,0.08876799543698628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,32,balanced,0.0988106628259023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,32,balanced,0.1288159986337026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,32,balanced,0.18447466691335043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,32,balanced,0.19613866011301676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,32,balanced,0.19670933485031128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,32,balanced,0.19777067502339682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,32,balanced,0.20975999037424722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,32,balanced,0.20481600364049277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,32,balanced,0.19534399112065634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,32,balanced,0.20459733406702676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,32,balanced,0.19822933276494345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,32,balanced,0.20538665850957236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,32,balanced,0.21555733680725098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,32,balanced,0.20302400986353555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,32,balanced,0.21146132548650107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,32,balanced,0.2127573291460673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,32,balanced,0.2124533255894979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,32,balanced,0.06393600006898244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,32,balanced,0.2164106567700704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,32,balanced,0.06513066589832306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,32,balanced,0.06373866895834605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,32,balanced,0.07950933277606964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,32,balanced,0.090938667456309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,32,balanced,0.23775466283162436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,32,balanced,0.12982933719952902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,32,balanced,0.12929067015647888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,32,balanced,0.12313066919644673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,32,balanced,0.13685866196950278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,32,balanced,0.1295840044816335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,32,balanced,0.12689066926638284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,32,balanced,0.2594613234202067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,32,balanced,0.1288479963938395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,32,balanced,0.13447466492652893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,32,balanced,0.13454399506251016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,32,balanced,0.13296533624331155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,32,balanced,0.13920533657073975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,32,balanced,0.38340266545613605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,32,balanced,0.14155200123786926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,32,balanced,0.1499679982662201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,32,balanced,0.16110933820406595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,32,balanced,0.17847466468811035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,32,balanced,0.39395201206207275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,32,balanced,0.1923253337542216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,32,balanced,0.22859734296798706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,32,balanced,0.26571200291315716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,32,balanced,0.5413920084635416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,32,balanced,0.37323200702667236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,32,balanced,0.06267199913660686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,32,balanced,0.06333866715431213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,32,balanced,0.06648000081380208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,32,balanced,0.07663466533025105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,32,balanced,0.09248000383377075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,32,balanced,0.14713066816329956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,32,balanced,0.13890666762987772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,32,balanced,0.43406399091084796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,32,balanced,0.1344053347905477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,32,balanced,0.13423466682434082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,32,balanced,0.13136000434557596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,32,balanced,0.12433600425720215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,32,balanced,0.1276533305644989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,32,balanced,0.13285866379737854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,32,balanced,0.04167466859022776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,32,balanced,0.04342400034268697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,32,balanced,0.13194132844607034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,32,balanced,0.043562665581703186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,32,balanced,0.6678133010864258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,32,balanced,0.13699199755986533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,32,balanced,0.04260799785455068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,32,balanced,0.14100266496340433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,32,balanced,0.046570668617884316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,32,balanced,0.6061760187149048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,32,balanced,0.05755199988683065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,32,balanced,0.14070399602254233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,32,balanced,0.07226133346557617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,32,balanced,0.07057066758473714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,32,balanced,0.13912000258763632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,32,balanced,0.09387733538945515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,32,balanced,0.1509066621462504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,32,balanced,0.09398933251698811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,32,balanced,0.09505066275596619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,32,balanced,0.10728533069292705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,32,balanced,0.16169066230456033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,32,balanced,0.10563733180363973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,32,balanced,0.1013759970664978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,32,balanced,0.16588800152142844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,32,balanced,0.1250879963239034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,32,balanced,0.11924266815185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,32,balanced,0.7726720174153646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,32,balanced,0.1107306679089864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,32,balanced,0.18715733289718628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,32,balanced,0.13084800044695535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,32,balanced,0.15292267004648843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,32,balanced,0.8132479985555013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,32,balanced,0.2134079933166504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,32,balanced,0.18549333016077676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,32,balanced,0.23174399137496948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,32,balanced,0.3120959997177124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,32,balanced,0.3036213318506877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,32,balanced,0.9421599706013998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,32,balanced,0.38199468453725177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,32,balanced,0.341541330019633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,32,balanced,0.5262240171432495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,32,balanced,0.47204267978668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,32,balanced,0.6938772996266683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,32,balanced,1.2349546750386555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,32,balanced,1.0506079991658528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,32,balanced,1.453445275624593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,32,balanced,0.6022239923477173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,32,balanced,1.3628427187601726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,32,balanced,0.7297759850819906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,32,balanced,1.7102293968200684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,32,balanced,1.1156960328420003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,32,balanced,2.6811145146687827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,32,balanced,2.860714594523112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,32,balanced,2.398320039113363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,32,balanced,5.29145081837972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,32,balanced,2.188586711883545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,32,power_law_1.01,0.20465919971466065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,32,power_law_1.01,0.18254719972610473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,32,power_law_1.01,0.15421439409255983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,32,power_law_1.01,0.12955520153045655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,32,power_law_1.01,0.11352959871292115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,32,power_law_1.01,0.13191039562225343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,32,power_law_1.01,0.13409279584884642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,32,power_law_1.01,0.11760640144348145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,32,power_law_1.01,0.0861631989479065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,32,power_law_1.01,0.16286079883575438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,32,power_law_1.01,0.10814080238342286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,32,power_law_1.01,0.18723200559616088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,32,power_law_1.01,0.11567360162734985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,32,power_law_1.01,0.17886719703674317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,32,power_law_1.01,0.12094719409942627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,32,power_law_1.01,0.19299839735031127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,32,power_law_1.01,0.1247488021850586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,32,power_law_1.01,0.18818559646606445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,32,power_law_1.01,0.12250880002975464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,32,power_law_1.01,0.1877568006515503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,32,power_law_1.01,0.13264000415802002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,32,power_law_1.01,0.1915519952774048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,32,power_law_1.01,0.12420480251312256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,32,power_law_1.01,0.19325439929962157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,32,power_law_1.01,0.1320512056350708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,32,power_law_1.01,0.18079999685287476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,32,power_law_1.01,0.12716799974441528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,32,power_law_1.01,0.19829119443893434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,32,power_law_1.01,0.1286911964416504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,32,power_law_1.01,0.2010495901107788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,32,power_law_1.01,0.13689600229263305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,32,power_law_1.01,0.1962175965309143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,32,power_law_1.01,0.13867520093917846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,32,power_law_1.01,0.2317888021469116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,32,power_law_1.01,0.13704320192337036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,32,power_law_1.01,0.22405760288238524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,32,power_law_1.01,0.15914880037307738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,32,power_law_1.01,0.2722176074981689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,32,power_law_1.01,0.18161920309066773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,32,power_law_1.01,0.30716800689697266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,32,power_law_1.01,0.20832641124725343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,32,power_law_1.01,0.40488319396972655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,32,power_law_1.01,0.2475519895553589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,32,power_law_1.01,0.42892160415649416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,32,power_law_1.01,0.3041984081268311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,32,power_law_1.01,0.5427264213562012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,32,power_law_1.01,0.1431488037109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,32,power_law_1.01,0.3642303943634033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,32,power_law_1.01,0.7161664009094239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,32,power_law_1.01,0.1323456048965454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,32,power_law_1.01,0.5146495819091796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,32,power_law_1.01,0.08986240029335021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,32,power_law_1.01,0.9036735534667969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,32,power_law_1.01,0.6115392208099365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,32,power_law_1.01,0.08824319839477539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,32,power_law_1.01,1.352025604248047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,32,power_law_1.01,0.04973439872264862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,32,power_law_1.01,0.10478719472885131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,32,power_law_1.01,0.8604607582092285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,32,power_law_1.01,1.776358413696289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,32,power_law_1.01,0.04940800070762634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,32,power_law_1.01,0.11327359676361085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,32,power_law_1.01,1.086207962036133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,32,power_law_1.01,2.217215919494629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,32,power_law_1.01,0.1205183982849121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,32,power_law_1.01,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,32,power_law_1.01,1.4401791572570801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,32,power_law_1.01,0.128985595703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,32,power_law_1.01,4.786336135864258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,32,power_law_1.01,2.026668739318848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,32,power_law_1.01,0.12358399629592895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,32,power_law_1.01,0.12524800300598143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,32,power_law_1.01,4.077702331542969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,32,power_law_1.01,0.12372479438781739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,32,power_law_1.01,0.13133440017700196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,32,power_law_1.01,0.12249599695205689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,32,power_law_1.01,0.13225599527359008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,32,power_law_1.01,0.13365119695663452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,32,power_law_1.01,0.13633279800415038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,32,power_law_1.01,0.13430399894714357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,32,power_law_1.01,0.15163520574569703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.01,0.15797760486602783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.01,0.18978559970855713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.01,0.22641921043395996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.01,0.2688960075378418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.01,0.3099391937255859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.01,0.41790080070495605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.01,0.5351359844207764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.01,0.7561408042907715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.01,0.9593088150024414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,32,power_law_1.01,1.102291202545166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,32,power_law_1.01,1.890278434753418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,32,power_law_1.01,3.208492660522461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,32,power_law_1.2,0.19228800535202026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,32,power_law_1.2,0.1832128047943115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,32,power_law_1.2,0.09377279877662659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,32,power_law_1.2,0.1223296046257019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,32,power_law_1.2,0.1343999981880188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,32,power_law_1.2,0.13364479541778565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,32,power_law_1.2,0.155731201171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,32,power_law_1.2,0.1355072021484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,32,power_law_1.2,0.1828223943710327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,32,power_law_1.2,0.069760000705719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,32,power_law_1.2,0.18609919548034667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,32,power_law_1.2,0.18986239433288574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,32,power_law_1.2,0.08826879858970642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,32,power_law_1.2,0.18583680391311647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,32,power_law_1.2,0.09284480214118958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,32,power_law_1.2,0.1841920018196106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,32,power_law_1.2,0.11509120464324951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,32,power_law_1.2,0.19922560453414917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,32,power_law_1.2,0.11547520160675048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,32,power_law_1.2,0.192467200756073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,32,power_law_1.2,0.1279744029045105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,32,power_law_1.2,0.18021759986877442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,32,power_law_1.2,0.12438399791717529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,32,power_law_1.2,0.1992640018463135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,32,power_law_1.2,0.12676479816436767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,32,power_law_1.2,0.19189759492874145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,32,power_law_1.01,0.03978880047798157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,32,power_law_1.2,0.1257024049758911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,32,power_law_1.2,0.22421119213104249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,32,power_law_1.01,0.04325119853019714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,32,power_law_1.2,0.13121919631958007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,32,power_law_1.2,0.24333438873291016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,32,power_law_1.01,0.049132800102233885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,32,power_law_1.2,0.1253376007080078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,32,power_law_1.2,0.2686271905899048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,32,power_law_1.01,0.07683839797973632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,32,power_law_1.2,0.1347584009170532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,32,power_law_1.2,0.30437119007110597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,32,power_law_1.01,0.07400959730148315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,32,power_law_1.2,0.13757439851760864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,32,power_law_1.2,0.3368383884429932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,32,power_law_1.01,0.09446399807929992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,32,power_law_1.2,0.13501440286636351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,32,power_law_1.2,0.14225280284881592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,32,power_law_1.2,0.1431167960166931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,32,power_law_1.2,0.4550144195556641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,32,power_law_1.2,0.14231040477752685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,32,power_law_1.2,0.16620160341262818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,32,power_law_1.2,0.5290112018585205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,32,power_law_1.2,0.08206080198287964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,32,power_law_1.2,0.19274239540100097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,32,power_law_1.2,0.6392767906188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,32,power_law_1.2,0.0880832016468048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,32,power_law_1.2,0.22391040325164796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,32,power_law_1.2,0.905452823638916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,32,power_law_1.2,0.10529279708862305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,32,power_law_1.2,0.259500789642334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,32,power_law_1.2,1.3804032325744628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,32,power_law_1.2,0.10848640203475952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,32,power_law_1.2,0.3401407957077026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,32,power_law_1.2,1.7239999771118164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,32,power_law_1.2,0.11646720170974731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,32,power_law_1.2,0.4100224018096924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,32,power_law_1.2,2.3124416351318358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,32,power_law_1.2,0.11955840587615967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,32,power_law_1.2,0.572979211807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,32,power_law_1.2,0.12369279861450196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,32,power_law_1.2,3.2973377227783205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,32,power_law_1.2,0.7582079887390136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,32,power_law_1.2,0.12984960079193114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,32,power_law_1.2,6.164409637451172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,32,power_law_1.2,1.0156671524047851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,32,power_law_1.2,0.13315839767456056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,32,power_law_1.2,0.12765439748764038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,32,power_law_1.2,1.5272192001342773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,32,power_law_1.2,0.13087999820709229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,32,power_law_1.2,1.6223360061645509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,32,power_law_1.2,0.13334399461746216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,32,power_law_1.2,2.590291213989258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,32,power_law_1.2,0.13692159652709962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,32,power_law_1.2,0.13861759901046752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,32,power_law_1.2,5.4096832275390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,32,power_law_1.2,0.14228479862213134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,32,power_law_1.2,0.16020480394363404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.2,0.17132799625396727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.2,0.20741119384765624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.2,0.23463039398193358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.2,0.2910271883010864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.2,0.36913280487060546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.2,0.4845248222351074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.2,0.5881728172302246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.2,0.8195584297180176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.2,1.134438419342041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,32,power_law_1.2,1.4482687950134276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,32,power_law_1.2,1.9806463241577148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,32,power_law_1.2,4.181017684936523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,32,power_law_1.2,0.04819200038909912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,32,power_law_1.2,0.049420800805091855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,32,power_law_1.2,0.033964800834655764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,16,balanced,0.0668213317791621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,32,power_law_1.2,0.04206719994544983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,16,balanced,0.06879466772079468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,16,balanced,0.07739733159542084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,32,power_law_1.2,0.04389120042324066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,16,balanced,0.09464533130327861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,32,power_law_1.01,0.09303680062294006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,16,balanced,0.12446932991345723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,32,power_law_1.2,0.04897280037403107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,16,balanced,0.18705600500106812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,32,power_law_1.01,0.09407359957695008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,32,power_law_1.2,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,16,balanced,0.1789813240369161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,32,power_law_1.01,0.10951679944992065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,32,power_law_1.2,0.06965759992599488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,16,balanced,0.19113600254058838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,32,power_law_1.01,0.10595200061798096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,16,balanced,0.19908267259597778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,32,power_law_1.2,0.0925055980682373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,16,balanced,0.1939199964205424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,32,power_law_1.01,0.10071680545806885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,32,power_law_1.2,0.09374079704284669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,16,balanced,0.18594666322072348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,32,power_law_1.01,0.09815679788589478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,32,power_law_1.2,0.09197440147399902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,16,balanced,0.18149334192276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,32,power_law_1.01,0.09994879961013795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,32,power_law_1.2,0.11046400070190429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,16,balanced,0.18992000818252563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,32,power_law_1.01,0.10203520059585572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,32,power_law_1.2,0.11431039571762085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,16,balanced,0.18437866369883218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,32,power_law_1.2,0.10327039957046509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,32,power_law_1.01,0.10539519786834717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,16,balanced,0.196560005346934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,32,power_law_1.2,0.10045440196990967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,16,balanced,0.19126399358113608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.01,0.10940159559249878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,16,balanced,0.1929546594619751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,32,power_law_1.2,0.10182399749755859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.01,0.11703039407730102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,32,power_law_1.2,0.10611840486526489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.01,0.13731839656829833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,16,balanced,0.20417600870132446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,32,power_law_1.2,0.10350079536437988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.01,0.17614079713821412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,16,balanced,0.19976532459259033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.2,0.1133504033088684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.01,0.20694398880004883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.2,0.12698880434036255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.01,0.3085119962692261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,16,balanced,0.2109653353691101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,16,balanced,0.054330666859944664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.2,0.14102400541305543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.01,0.3911871910095215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,16,balanced,0.05548266569773356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,16,balanced,0.061568001906077065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,16,balanced,0.22137600183486938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.2,0.18596479892730713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.01,0.5785664081573486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,16,balanced,0.07310933371384938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.2,0.21685760021209716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,16,balanced,0.09268266956011455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.01,0.7502399921417237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,16,balanced,0.1325973371664683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.2,0.34232320785522463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,16,balanced,0.2458933393160502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,32,power_law_1.01,0.952518367767334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,16,balanced,0.12748799721399942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.2,0.4267136096954346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,16,balanced,0.12470933794975281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,32,power_law_1.01,1.445024013519287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,16,balanced,0.12578133742014566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.2,0.6455103874206543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,16,balanced,0.12787200013796488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,16,balanced,0.2704640030860901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,32,power_law_1.01,2.8874496459960937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.2,0.8323583602905273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,16,balanced,0.12285332878430684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,16,balanced,0.12379733721415202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,32,power_law_1.2,1.1079423904418946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,16,balanced,0.12566933035850525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,32,power_law_1.2,1.6199039459228515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,16,balanced,0.12841066718101501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,16,balanced,0.38556798299153644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,16,balanced,0.13521066308021545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,32,power_law_1.2,3.7457088470458983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,16,balanced,0.13587199648221335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,16,balanced,0.1379146675268809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,16,balanced,0.15565866231918335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,16,balanced,0.4161386489868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,16,balanced,0.15759467085202536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,16,balanced,0.1827359994252523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,16,balanced,0.19820799430211386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,16,balanced,0.5718666712443033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,16,balanced,0.23958933353424072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,16,balanced,0.27992000182469684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,16,balanced,0.7060853640238444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,16,balanced,0.3880106608072917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,16,balanced,0.4610613187154134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,16,balanced,0.8608799775441488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,16,balanced,0.6449813445409139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,16,balanced,0.825429360071818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,16,balanced,1.3148319721221924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,16,balanced,1.016213337580363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,16,balanced,0.054842665791511536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,16,balanced,0.059157331784566246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,16,balanced,0.06316266457239787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,16,balanced,0.07778666913509369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,16,balanced,0.09488532940546672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,16,balanced,0.1381066640218099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,16,balanced,1.567776044209798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,16,balanced,0.13428266843159994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,16,balanced,0.13426132996877035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,16,balanced,0.025392000873883564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,16,balanced,0.13379733761151633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,16,balanced,0.02628266563018163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,16,balanced,0.1353386640548706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,16,balanced,2.5591519673665366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,16,balanced,0.02693866689999898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,16,balanced,0.1345919966697693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,16,balanced,0.029322666426499683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,16,balanced,0.133242666721344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,16,balanced,0.0412266676624616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,16,balanced,0.055498664577802025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,16,balanced,0.12870933612187704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,16,balanced,0.06284266710281372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,16,balanced,0.13487466176350912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,16,balanced,0.060533334811528526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,16,balanced,0.13318933049837747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,16,balanced,0.084389328956604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,16,balanced,0.14298666516939798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,16,balanced,0.08402666449546814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,16,balanced,0.07619733115037282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,16,balanced,0.13983466227849325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,16,balanced,0.11923733353614807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,16,balanced,0.1511840025583903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,16,balanced,0.12050666411717732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,16,balanced,0.12241599957148235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,16,balanced,0.15359999736150107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,16,balanced,0.1281706690788269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,16,balanced,0.16951467593510947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,16,balanced,0.12125333150227864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,16,balanced,0.12730133533477783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,16,balanced,0.17442667484283447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,16,balanced,0.13587199648221335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,16,balanced,0.14451199769973755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,16,balanced,0.20114666223526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,16,balanced,0.16471999883651733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,16,balanced,0.1909653345743815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,16,balanced,0.2286133368810018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,16,balanced,0.23715200026830038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,16,balanced,0.32387200991312665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,16,balanced,0.29226134220759076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,16,balanced,0.3688746690750122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,16,balanced,0.3714826504389445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,16,balanced,3.071536064147949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,16,balanced,0.46407465140024823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,16,balanced,0.5171093146006266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,16,balanced,0.7192426522572836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,16,balanced,0.6611146529515585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,16,balanced,0.9204373359680176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,16,balanced,0.8099573453267416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,16,balanced,1.164453347524007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,16,balanced,1.242517312367757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,16,power_law_1.01,0.12511999607086183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,16,power_law_1.01,0.19911680221557618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,16,balanced,1.7936959266662598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,16,power_law_1.01,0.08298239707946778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,16,power_law_1.01,0.1033087968826294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,16,power_law_1.01,0.12182400226593018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,16,power_law_1.01,0.13954559564590455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,16,power_law_1.01,0.16209919452667237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,16,power_law_1.01,0.176800000667572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,16,power_law_1.01,0.17519999742507936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,16,power_law_1.01,0.17742079496383667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,16,power_law_1.01,0.17464319467544556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,16,power_law_1.01,0.18528640270233154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,16,power_law_1.01,0.17882879972457885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,16,power_law_1.01,0.18697600364685057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,16,power_law_1.01,0.18474240303039552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,16,power_law_1.01,0.19070080518722535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,16,power_law_1.01,0.18360320329666138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,16,power_law_1.01,0.20126080513000488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,16,power_law_1.01,0.2203968048095703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,16,power_law_1.01,0.2631488084793091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,16,power_law_1.01,0.2862400054931641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,16,power_law_1.01,0.35401599407196044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,16,power_law_1.01,0.3955391883850098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,16,power_law_1.01,0.5370880126953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,16,power_law_1.01,0.6360447883605957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,16,power_law_1.01,0.8901503562927247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,16,power_law_1.01,1.1943552017211914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,16,power_law_1.01,1.534425640106201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,16,power_law_1.01,0.08985599875450134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,16,balanced,2.429157257080078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,16,power_law_1.01,2.1895168304443358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,16,power_law_1.01,0.12554880380630493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,16,power_law_1.01,0.07849599719047547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,16,power_law_1.01,3.8635585784912108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,16,power_law_1.01,0.07448959946632386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,16,power_law_1.01,0.09207680225372314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,16,power_law_1.01,0.09704319834709167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,16,power_law_1.01,0.11247999668121338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,16,balanced,3.556346575419108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,16,power_law_1.01,0.11614079475402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,16,power_law_1.01,0.11763839721679688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,16,power_law_1.01,0.11771520376205444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,16,power_law_1.01,0.12261760234832764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,16,power_law_1.01,0.12264959812164307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,16,power_law_1.01,0.12333439588546753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,16,power_law_1.01,0.12693120241165162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,16,power_law_1.01,0.135315203666687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,16,power_law_1.01,0.13463040590286254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,16,power_law_1.01,0.13946880102157594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,16,power_law_1.01,0.1551103949546814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,16,power_law_1.01,0.168614399433136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,16,power_law_1.01,0.1994047999382019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,16,power_law_1.01,0.22740480899810792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,16,power_law_1.01,0.29690239429473875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,16,power_law_1.01,0.3509376049041748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,16,power_law_1.01,0.47951359748840333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,16,power_law_1.01,0.5961023807525635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,16,power_law_1.01,0.8030847549438477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,16,power_law_1.01,1.0383423805236816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,16,power_law_1.01,1.3452223777770995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,16,power_law_1.01,2.055059242248535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,16,power_law_1.01,4.095206451416016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,16,power_law_1.01,0.027801600098609925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,16,power_law_1.01,0.04845440089702606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,16,power_law_1.01,0.10227199792861938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,16,power_law_1.01,0.024742400646209715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,16,power_law_1.01,0.14494080543518068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,16,power_law_1.01,0.02847360074520111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,16,power_law_1.01,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,16,power_law_1.01,0.03845120072364807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,16,power_law_1.01,0.08293120265007019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,16,power_law_1.01,0.0418368011713028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,16,power_law_1.01,0.09773439764976502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,16,power_law_1.01,0.05018879771232605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,16,power_law_1.01,0.10069760084152221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,16,power_law_1.01,0.05496960282325745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,16,power_law_1.01,0.1249343991279602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,16,power_law_1.2,0.12161920070648194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,16,power_law_1.01,0.0691968023777008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,16,power_law_1.01,0.1208575963973999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,16,power_law_1.2,0.1960576057434082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,16,power_law_1.01,0.07121919989585876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,16,power_law_1.01,0.12318719625473022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,16,power_law_1.2,0.08133119940757752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,16,power_law_1.01,0.06513280272483826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,16,power_law_1.01,0.12355200052261353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,16,power_law_1.01,0.11946239471435546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,16,power_law_1.01,0.09292160272598267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,16,power_law_1.2,0.10030080080032348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,16,power_law_1.01,0.12218879461288452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,16,power_law_1.01,0.09316480159759521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,16,power_law_1.2,0.11735039949417114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,16,power_law_1.01,0.12857600450515747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,16,power_law_1.01,0.09386240243911743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,16,power_law_1.2,0.12660479545593262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,16,power_law_1.01,0.10274560451507568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,16,power_law_1.01,0.1247488021850586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,16,power_law_1.2,0.15986560583114623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,16,power_law_1.01,0.10488320589065551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,16,power_law_1.01,0.13065600395202637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,16,power_law_1.2,0.16501760482788086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,16,power_law_1.01,0.10743039846420288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,16,power_law_1.01,0.13185919523239137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,16,power_law_1.2,0.17032959461212158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,16,power_law_1.01,0.10180480480194092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,16,power_law_1.01,0.14142719507217408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,16,power_law_1.2,0.18138240575790404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.01,0.1093567967414856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,16,power_law_1.01,0.15045119524002076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.01,0.12184319496154786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.01,0.16090240478515624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,16,power_law_1.2,0.1747648000717163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.01,0.1265727996826172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.01,0.18466559648513795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,16,power_law_1.2,0.1768064022064209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.01,0.1595136046409607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.01,0.21595520973205568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,16,power_law_1.2,0.17897599935531616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.01,0.26965761184692383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.01,0.18981120586395264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,16,power_law_1.2,0.18238719701766967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.01,0.32792320251464846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.01,0.2551039934158325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,16,power_law_1.2,0.18336000442504882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.01,0.42466559410095217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.01,0.323635196685791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,16,power_law_1.2,0.18705919981002808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.01,0.5268159866333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.01,0.4466303825378418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,16,power_law_1.2,0.20064640045166016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.01,0.7063295841217041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.01,0.5814911842346191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,16,power_law_1.2,0.21809279918670654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.01,0.9553728103637695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,16,power_law_1.01,0.7352831840515137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,16,power_law_1.2,0.24554240703582764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,16,power_law_1.2,0.09066240191459655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,16,power_law_1.01,1.0905728340148926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,16,power_law_1.01,1.0841983795166015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,16,power_law_1.2,0.2625024080276489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,16,power_law_1.2,0.12539520263671874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,16,power_law_1.01,1.8175935745239258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,16,power_law_1.2,0.31348481178283694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,16,power_law_1.01,2.179148864746094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,16,power_law_1.2,0.06963840126991272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,16,power_law_1.2,0.3862528085708618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,16,power_law_1.01,3.5585407257080077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,16,power_law_1.2,0.07741439938545228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,16,power_law_1.2,0.08584960103034973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,16,power_law_1.2,0.48415360450744627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,16,power_law_1.2,0.0932864010334015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,16,power_law_1.2,0.6211904048919678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,16,power_law_1.2,0.1150015950202942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,16,power_law_1.2,0.801689624786377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,16,power_law_1.2,0.11335680484771729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,16,power_law_1.2,1.0078399658203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,16,power_law_1.2,0.11633280515670777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,16,power_law_1.2,1.3127103805541993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,16,power_law_1.2,0.11342719793319703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,16,power_law_1.2,0.12328959703445434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,16,power_law_1.2,1.9215295791625977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,16,power_law_1.2,0.12277760505676269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,16,power_law_1.2,2.625913619995117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,16,power_law_1.2,0.12299519777297974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,16,power_law_1.2,6.047321701049805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,16,power_law_1.2,0.12394239902496337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,16,power_law_1.2,0.13654400110244752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,16,power_law_1.2,0.1389248013496399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,16,power_law_1.2,0.15073280334472655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,16,power_law_1.2,0.1603327989578247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,16,power_law_1.2,0.18241280317306519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,16,power_law_1.2,0.21066880226135254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,16,power_law_1.2,0.2465343952178955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,16,power_law_1.2,0.30958080291748047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,16,power_law_1.2,0.38035199642181394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,16,power_law_1.2,0.5404287815093994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,16,power_law_1.2,0.6801983833312988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,16,power_law_1.2,0.9569600105285645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,16,power_law_1.2,1.2637311935424804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,16,power_law_1.2,1.58919038772583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,16,power_law_1.2,2.156300735473633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,16,power_law_1.2,4.83447036743164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,16,power_law_1.2,0.10364799499511719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,16,power_law_1.2,0.13587839603424073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,16,power_law_1.2,0.06618239879608154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,16,power_law_1.2,0.027212798595428467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,16,power_law_1.2,0.08665599822998046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,16,power_law_1.2,0.04943360090255737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,16,power_law_1.2,0.09007359743118286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,16,power_law_1.2,0.026553601026535034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,16,power_law_1.2,0.10209280252456665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,16,power_law_1.2,0.028799998760223388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,16,power_law_1.2,0.11400959491729737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,16,power_law_1.2,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,16,power_law_1.2,0.12113920450210572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,16,power_law_1.2,0.0398719996213913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,16,power_law_1.2,0.1229375958442688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,16,power_law_1.2,0.05074560046195984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,16,power_law_1.2,0.1176576018333435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,16,power_law_1.2,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,16,power_law_1.2,0.12789759635925294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,16,power_law_1.2,0.06442239880561829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,16,power_law_1.2,0.1260480046272278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,16,power_law_1.2,0.06309120059013366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,16,power_law_1.2,0.12870399951934813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,16,power_law_1.2,0.06755840182304382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,16,power_law_1.2,0.13144960403442382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,16,power_law_1.2,0.09342719912528992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,16,power_law_1.2,0.13717759847640992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,8,balanced,0.05900266766548157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,16,power_law_1.2,0.09536640048027038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,8,balanced,0.0625600020090739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,16,power_law_1.2,0.13841279745101928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,16,power_law_1.2,0.09527040123939515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,8,balanced,0.07195200026035309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,16,power_law_1.2,0.1410688042640686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,16,power_law_1.2,0.1055616021156311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,8,balanced,0.09443733096122742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,8,balanced,0.1218826671441396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,16,power_law_1.2,0.15861120223999023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,16,power_law_1.2,0.10088959932327271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,8,balanced,0.19457600514094034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.2,0.16733440160751342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,16,power_law_1.2,0.10329600572586059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,8,balanced,0.1886133352915446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.2,0.197708797454834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,8,balanced,0.19046932458877563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,16,power_law_1.2,0.10228480100631714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,8,balanced,0.17938133080800375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.2,0.2369920015335083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.2,0.11502079963684082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,8,balanced,0.19301867485046387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.2,0.28573439121246336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.2,0.11752959489822387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,8,balanced,0.1766080061594645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.2,0.35571839809417727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,8,balanced,0.1736639936765035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.2,0.1266816020011902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,8,balanced,0.18360000848770142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.2,0.5034687995910645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.2,0.16659200191497803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,8,balanced,0.1795253356297811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.2,0.5420032024383545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.2,0.18799359798431398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,8,balanced,0.1875093380610148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.2,0.7689023971557617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.2,0.2642944097518921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,8,balanced,0.19138665994008383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.2,1.0425408363342286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.2,0.3449728012084961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,8,balanced,0.1865760087966919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,16,power_law_1.2,1.3749376296997071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.2,0.47054080963134765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,8,balanced,0.19941866397857666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.2,0.6437312126159668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,16,power_law_1.2,2.0253055572509764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,16,power_law_1.2,0.8170623779296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,8,balanced,0.20387732982635498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,16,power_law_1.2,4.437529754638672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,8,balanced,0.05049600203831991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,16,power_law_1.2,1.2761024475097655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,8,balanced,0.055760001142819725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,8,balanced,0.22478934129079184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,8,balanced,0.0595413347085317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,16,power_law_1.2,2.5856000900268556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,8,balanced,0.07286400099595387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,8,balanced,0.09006399909655254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,8,balanced,0.23321066300074259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,8,balanced,0.12874666849772134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,8,balanced,0.12753599882125854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,8,balanced,0.12657599647839865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,8,balanced,0.2659839987754822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,8,balanced,0.12845333417256674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,8,balanced,0.12877333164215088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,8,balanced,0.12406399846076965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,8,balanced,0.12735999623934427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,8,balanced,0.2966933250427246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,8,balanced,0.1336373289426168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,8,balanced,0.13099732995033264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,8,balanced,0.1376213332017263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,8,balanced,0.1402773360411326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,8,balanced,0.14517333110173544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,8,balanced,0.4168479839960734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,8,balanced,0.15761599938074747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,8,balanced,0.16987733046213785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,8,balanced,0.19450666507085165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,8,balanced,0.46850132942199707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,8,balanced,0.21474132935206094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,8,balanced,0.2635093331336975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,8,balanced,0.652890682220459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,8,balanced,0.30828799804051715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,8,balanced,0.436245322227478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,8,balanced,0.8116426467895508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,8,balanced,0.5232693354288737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,8,balanced,0.7426026662190756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,8,balanced,0.9860533078511556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,8,balanced,0.9524319966634115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,8,balanced,1.5192426045735676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,8,balanced,1.164021333058675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,8,balanced,0.05403199791908264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,8,balanced,0.06097066899140676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,8,balanced,0.06308266520500183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,8,balanced,0.07578666508197784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,8,balanced,0.0236160010099411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,8,balanced,0.10525332887967427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,8,balanced,0.02480533222357432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,8,balanced,0.15375999609629312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,8,balanced,0.02796799937884013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,8,balanced,0.155130664507548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,8,balanced,0.028575999041398365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,8,balanced,0.14242666959762573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,8,balanced,0.041093334555625916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,8,balanced,0.14563199877738953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,8,balanced,0.05876266459623972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,8,balanced,0.13797332843144736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,8,balanced,0.05850133299827576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,8,balanced,0.13665067156155905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,8,balanced,0.061424002051353455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,8,balanced,0.14290133118629456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,8,balanced,0.0652213344971339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,8,balanced,0.06509333352247874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,8,balanced,0.14525333046913147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,8,balanced,0.06836266815662384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,8,balanced,0.1416266659895579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,8,balanced,0.0895146628220876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,8,balanced,0.145087997118632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,8,balanced,0.0813920001188914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,8,balanced,0.154831995566686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,8,balanced,0.08878933389981587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,8,balanced,0.1509760022163391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,8,balanced,0.12500266234079996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,8,balanced,0.16460800170898438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,8,balanced,0.1251306633154551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,8,balanced,1.7974826494852703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,8,balanced,0.13051199913024902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,8,balanced,0.16473066806793213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,8,balanced,0.14194132884343466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,8,balanced,0.18059200048446655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,8,balanced,0.15742400288581848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,8,balanced,0.1958720088005066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,8,balanced,0.17047999302546182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,8,balanced,0.22975466648737589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,8,balanced,0.1819093426068624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,8,balanced,0.213754673798879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,8,balanced,0.26497066020965576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,8,balanced,2.964954694112142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,8,balanced,0.24775999784469604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,8,balanced,0.3799999952316284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,8,balanced,0.30906132857004803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,8,balanced,0.4421333471934001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,8,balanced,0.3714880148569743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,8,balanced,0.6272266705830892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,8,balanced,0.5728000005086263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,8,balanced,0.806821346282959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,8,balanced,3.524933179219564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,8,balanced,0.71396803855896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,8,balanced,0.9786346753438314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,8,balanced,0.9291199843088785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,8,balanced,1.5231626828511555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,8,balanced,1.4017386436462402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,8,power_law_1.01,0.09547520279884339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,8,power_law_1.01,0.11967999935150146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,8,power_law_1.01,0.08991360068321227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,8,power_law_1.01,0.09262080192565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,8,power_law_1.01,0.11079679727554322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,8,power_law_1.01,0.12321280241012574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,8,power_law_1.01,0.15267200469970704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,8,balanced,2.9709653854370117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,8,power_law_1.01,0.16238720417022706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,8,power_law_1.01,0.1689471960067749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,8,power_law_1.01,0.1688447952270508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,8,power_law_1.01,0.17142399549484252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,8,power_law_1.01,0.1744704008102417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,8,power_law_1.01,0.1742143988609314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,8,balanced,2.763728141784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,8,power_law_1.01,0.17570559978485106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,8,power_law_1.01,0.18128639459609985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,8,power_law_1.01,0.18380160331726075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,8,power_law_1.01,0.1903807997703552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,8,power_law_1.01,0.20241920948028563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,8,power_law_1.01,0.2188096046447754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,8,power_law_1.01,0.2565119981765747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,8,power_law_1.01,0.274835205078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,8,power_law_1.01,0.33408000469207766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,8,power_law_1.01,0.40032000541687013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,8,power_law_1.01,0.07128959894180298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,8,power_law_1.01,0.5355455875396729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,8,power_law_1.01,0.09482880234718323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,8,power_law_1.01,0.07120000123977661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,8,power_law_1.01,0.6786367893218994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,8,power_law_1.01,0.0716480016708374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,8,power_law_1.01,0.8699071884155274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,8,power_law_1.01,0.08568959832191467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,8,power_law_1.01,1.0943743705749511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,8,power_law_1.01,0.09715840220451355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,8,power_law_1.01,1.4475839614868165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,8,power_law_1.01,0.11281280517578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,8,power_law_1.01,2.0314048767089843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,8,power_law_1.01,0.11932159662246704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,8,power_law_1.01,3.976607894897461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,8,power_law_1.01,0.11754239797592163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,8,power_law_1.01,0.12008960247039795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,8,power_law_1.01,0.12079360485076904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,8,power_law_1.01,0.12439039945602418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,8,power_law_1.01,0.12400000095367432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,8,power_law_1.01,0.12810879945755005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,8,power_law_1.01,0.1341055989265442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,8,power_law_1.01,0.1412160038948059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,8,power_law_1.01,0.1453376054763794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,8,power_law_1.01,0.16303999423980714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,8,power_law_1.01,0.18091520071029663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,8,power_law_1.01,0.21199359893798828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,8,power_law_1.01,0.23972480297088622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,8,power_law_1.01,0.306662392616272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,8,power_law_1.01,0.3693887948989868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,8,power_law_1.01,0.5167295932769775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,8,power_law_1.01,0.6274752140045166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,8,power_law_1.01,0.881772804260254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,8,power_law_1.01,1.1224575996398927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,8,power_law_1.01,1.3822527885437013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,8,power_law_1.01,2.2442880630493165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,8,power_law_1.01,4.399078369140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,8,power_law_1.01,0.08149759769439698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,8,power_law_1.01,0.10645120143890381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,8,power_law_1.01,0.07719039916992188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,8,power_law_1.01,0.08020480275154114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,8,power_law_1.01,0.09868159890174866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,8,power_law_1.01,0.10354559421539307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,8,power_law_1.01,0.12496639490127563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,8,power_law_1.01,0.023078399896621703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,8,power_law_1.01,0.13175679445266725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,8,power_law_1.01,0.028723201155662535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,8,power_law_1.01,0.12581119537353516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,8,power_law_1.01,0.02542079985141754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,8,power_law_1.01,0.13363840579986572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,8,power_law_1.01,0.02884480059146881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,8,power_law_1.01,0.1313279986381531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,8,power_law_1.01,0.034016001224517825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,8,power_law_1.01,0.13159040212631226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,8,power_law_1.01,0.038380798697471616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,8,power_law_1.01,0.13029760122299194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,8,power_law_1.01,0.04906240105628967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,8,power_law_1.01,0.13483519554138185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,8,power_law_1.01,0.05057920217514038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,8,power_law_1.01,0.1442944049835205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,8,power_law_1.01,0.052300798892974856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,8,power_law_1.01,0.1451647996902466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,8,power_law_1.01,0.056883198022842404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,8,power_law_1.01,0.14387199878692628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,8,power_law_1.01,0.05489280223846436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,8,power_law_1.01,0.16676479578018188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,8,power_law_1.01,0.06955519914627076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.01,0.17396479845046997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,8,power_law_1.01,0.06807039976119995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.01,0.2022144079208374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,8,power_law_1.01,0.06566399931907654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.01,0.22424960136413574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,8,power_law_1.01,0.09582719802856446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.01,0.2940351963043213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,8,power_law_1.01,0.09752960205078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.01,0.3474303960800171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,8,power_law_1.01,0.09920639991760254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.01,0.4533184051513672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,8,power_law_1.01,0.11404160261154175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.01,0.5451839923858642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.01,0.11376639604568481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,8,power_law_1.2,0.09429759979248047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.01,0.7575488090515137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.01,0.12566399574279785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,8,power_law_1.2,0.12205439805984497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.01,0.13137919902801515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.01,1.0690496444702149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,8,power_law_1.2,0.07591040134429931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.01,0.15980160236358643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,8,power_law_1.01,1.2483839988708496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,8,power_law_1.2,0.09201279878616334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.01,0.18448640108108522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,8,power_law_1.2,0.10428160429000854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,8,power_law_1.01,1.8790975570678712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.01,0.23429760932922364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,8,power_law_1.2,0.12037760019302368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,8,power_law_1.01,3.884588623046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.01,0.2882368087768555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,8,power_law_1.2,0.1533568024635315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.01,0.400051212310791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,8,power_law_1.2,0.1677888035774231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,8,power_law_1.2,0.16792960166931153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.01,0.5190591812133789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,8,power_law_1.2,0.1642624020576477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,8,power_law_1.01,0.6439040184020997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,8,power_law_1.2,0.17123199701309205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,8,power_law_1.01,0.9657919883728028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,8,power_law_1.2,0.17123199701309205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,8,power_law_1.2,0.17607040405273439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,8,power_law_1.01,1.8582656860351563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,8,power_law_1.2,0.18245760202407837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,8,power_law_1.2,0.18315520286560058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,8,power_law_1.2,0.19124480485916137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,8,power_law_1.2,0.194758403301239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,8,power_law_1.2,0.20469119548797607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,8,power_law_1.2,0.07245439887046815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,8,power_law_1.2,0.22633600234985352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,8,power_law_1.2,0.09534720182418824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,8,power_law_1.2,0.26467199325561525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,8,power_law_1.2,0.06322559714317322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,8,power_law_1.2,0.2897855997085571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,8,power_law_1.2,0.07854080200195312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,8,power_law_1.2,0.08170880079269409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,8,power_law_1.2,0.3619776010513306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,8,power_law_1.2,0.08870400190353393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,8,power_law_1.2,0.4323008060455322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,8,power_law_1.2,0.11569279432296753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,8,power_law_1.2,0.5630591869354248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,8,power_law_1.2,0.11548800468444824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,8,power_law_1.2,0.6652991771697998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,8,power_law_1.2,0.11938560009002686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,8,power_law_1.2,0.9920319557189942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,8,power_law_1.2,0.12076159715652465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,8,power_law_1.2,1.3552063941955566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,8,power_law_1.2,0.12006399631500245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,8,power_law_1.2,1.5740351676940918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,8,power_law_1.2,0.12278399467468262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,8,power_law_1.2,2.5038784027099608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,8,power_law_1.2,0.12649600505828856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,8,power_law_1.2,0.13151999711990356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,8,power_law_1.2,4.513183975219727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,8,power_law_1.2,0.13968640565872192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,8,power_law_1.2,0.14198399782180787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,8,power_law_1.2,0.15196800231933594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,8,power_law_1.2,0.16661759614944457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,8,power_law_1.2,0.18584959506988524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,8,power_law_1.2,0.22343039512634277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,8,power_law_1.2,0.24693119525909424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,8,power_law_1.2,0.32520959377288816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,8,power_law_1.2,0.3916095972061157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,8,power_law_1.2,0.547430419921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,8,power_law_1.2,0.6664000034332276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,8,power_law_1.2,0.9917632102966308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,8,power_law_1.2,1.1662912368774414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,8,power_law_1.2,1.462771224975586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,8,power_law_1.2,2.193497657775879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,8,power_law_1.2,5.745638275146485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,8,power_law_1.2,0.0824895977973938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,8,power_law_1.2,0.0997759997844696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,8,power_law_1.2,0.06899840235710145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,8,power_law_1.2,0.08172159790992736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,8,power_law_1.2,0.09571840167045594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,8,power_law_1.2,0.10371199846267701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,8,power_law_1.2,0.12788480520248413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,8,power_law_1.2,0.02213120013475418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,8,power_law_1.2,0.125382399559021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,8,power_law_1.2,0.027027198672294618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,8,power_law_1.2,0.12677119970321654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,8,power_law_1.2,0.02595840096473694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,8,power_law_1.2,0.12974079847335815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,8,power_law_1.2,0.02781440019607544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,8,power_law_1.2,0.133241605758667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,8,power_law_1.2,0.03198719918727875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,8,power_law_1.2,0.1331007957458496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,8,power_law_1.2,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,8,power_law_1.2,0.13288960456848145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,8,power_law_1.2,0.04831359982490539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,8,power_law_1.2,0.13694080114364623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,8,power_law_1.2,0.04864639937877655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,8,power_law_1.2,0.1430400013923645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,8,power_law_1.2,0.1530943989753723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,8,power_law_1.2,0.05170559883117676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,8,power_law_1.2,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,8,power_law_1.2,0.15207040309906006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,8,power_law_1.2,0.05392000079154968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,8,power_law_1.2,0.16666879653930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,8,power_law_1.2,0.06572800278663635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.2,0.18454400300979615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,8,power_law_1.2,0.06785920262336731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.2,0.21079039573669434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,8,power_law_1.2,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.2,0.23525760173797608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,8,power_law_1.2,0.09561600089073181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.2,0.29818880558013916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,8,power_law_1.2,0.09767680168151856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.2,0.36170239448547364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,8,power_law_1.2,0.09790080189704894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.2,0.517900800704956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,8,power_law_1.2,0.10381439924240113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.2,0.6040448188781739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.2,0.11148159503936768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.2,0.9092672348022461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.2,0.12755839824676513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.2,1.0866368293762207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.2,0.13502720594406128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,8,power_law_1.2,1.4599871635437012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,4,balanced,0.06258666515350342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.2,0.16239999532699584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,4,balanced,0.06606400012969971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,8,power_law_1.2,2.1377727508544924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.2,0.18509440422058104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,4,balanced,0.07746666669845581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,4,balanced,0.10017599662144978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.2,0.2527231931686401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,8,power_law_1.2,4.144268798828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,4,balanced,0.12354666988054912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.2,0.30548479557037356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,4,balanced,0.19478933016459146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,4,balanced,0.19243733088175455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.2,0.4213119983673096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,4,balanced,0.19474667310714722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.2,0.5346432209014893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,4,balanced,0.19165867567062378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,4,balanced,0.19634133577346802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,8,power_law_1.2,0.6528960227966308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,4,balanced,0.18799465894699097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,8,power_law_1.2,1.0773056030273438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,4,balanced,0.19193599621454874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,4,balanced,0.1904266675313314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,8,power_law_1.2,2.190239906311035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,4,balanced,0.19144533077875772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,4,balanced,0.1971199909845988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,4,balanced,0.1975946625073751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,4,balanced,0.20486400524775186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,4,balanced,0.22037333250045776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,4,balanced,0.054378668467203774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,4,balanced,0.22908266385396323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,4,balanced,0.05913066864013672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,4,balanced,0.06116800010204315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,4,balanced,0.08177599807580312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,4,balanced,0.24915200471878052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,4,balanced,0.10504000385602315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,4,balanced,0.14687466621398926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,4,balanced,0.14618666966756186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,4,balanced,0.26879467566808063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,4,balanced,0.14607466260592142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,4,balanced,0.14468800028165182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,4,balanced,0.1418719987074534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,4,balanced,0.3190986712773641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,4,balanced,0.14297599593798319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,4,balanced,0.14646400014559427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,4,balanced,0.14870400230089822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,4,balanced,0.15060800313949585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,4,balanced,0.1585599978764852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,4,balanced,0.3639306624730428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,4,balanced,0.16160533825556436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,4,balanced,0.1685333251953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,4,balanced,0.186298668384552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,4,balanced,0.19884800910949707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,4,balanced,0.5208959976832072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,4,balanced,0.23082667589187622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,4,balanced,0.25854400793711346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,4,balanced,0.6074133316675822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,4,balanced,0.3181546727816264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,4,balanced,0.377562681833903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,4,balanced,0.5438239971796671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,4,balanced,0.8489279747009277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,4,balanced,0.6664799849192301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,4,balanced,1.059397300084432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,4,balanced,0.9538186391194662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,4,balanced,1.2179733117421467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,4,balanced,1.2907360394795735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,4,balanced,0.060218666990598045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,4,balanced,0.06376533210277557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,4,balanced,0.07124799986680348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,4,balanced,1.4930613835652669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,4,balanced,0.08801066875457764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,4,balanced,0.11975466211636861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,4,balanced,0.17726399501164755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,4,balanced,0.17583467562993368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,4,balanced,0.16867733001708984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,4,balanced,0.1759679913520813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,4,balanced,0.17157334089279175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,4,balanced,0.16979199647903442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,4,balanced,0.02370133250951767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,4,balanced,0.1729653278986613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,4,balanced,0.024735999604066212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,4,balanced,0.17186667521794638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,4,balanced,0.026767998933792114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,4,balanced,0.17197332779566446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,4,balanced,0.03270400067170461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,4,balanced,0.03882133215665817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,4,balanced,0.17779199282328287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,4,balanced,0.056794668237368263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,4,balanced,0.18363734086354574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,4,balanced,0.060415998101234436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,4,balanced,0.18385599056879678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,4,balanced,0.06300800045331319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,4,balanced,1.9885867436726887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,4,balanced,0.06301866471767426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,4,balanced,0.19969600439071655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,4,balanced,0.06516799827416737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,4,balanced,0.06834133466084798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,4,balanced,0.2081706722577413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,4,balanced,0.07232533395290375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,4,balanced,0.07208533088366191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,4,balanced,2.3129706382751465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,4,balanced,0.23281067609786987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,4,balanced,0.07479999959468842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,4,balanced,0.09686932961146037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,4,balanced,0.2537173430124919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,4,balanced,0.09412800272305806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,4,balanced,0.10441600282986958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,4,balanced,0.30461867650349933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,4,balanced,0.14516266187032065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,4,balanced,0.15628266334533691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,4,balanced,0.3479893207550049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,4,balanced,0.1778293251991272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,4,balanced,0.2033066749572754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,4,balanced,0.5102186600367228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,4,balanced,0.21845867236455283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,4,balanced,0.6102453470230103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,4,balanced,0.24778133630752563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,4,balanced,0.30081599950790405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,4,balanced,0.8583146731058756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,4,balanced,0.3479199806849162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,4,balanced,1.1167946656545003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,4,balanced,3.8636693954467773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,4,balanced,4.547823905944824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,4,balanced,0.5350506703058878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,4,balanced,0.6513599952061971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,4,balanced,1.365023930867513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,4,balanced,0.8431946436564127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,4,balanced,2.1190932591756186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,4,balanced,1.2718613147735596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,4,balanced,4.164538701375325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,4,balanced,2.4876319567362466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,4,power_law_1.01,0.07921280264854431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,4,power_law_1.01,0.09887999892234803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,4,power_law_1.01,0.07473919987678528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,4,power_law_1.01,0.09642879962921143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,4,power_law_1.01,0.11003520488739013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,4,power_law_1.01,0.06724479794502258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,4,power_law_1.01,0.1254271984100342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,4,power_law_1.01,0.0767359972000122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,4,power_law_1.01,0.16644480228424072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,4,power_law_1.01,0.0694208025932312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,4,power_law_1.01,0.17000319957733154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,4,power_law_1.01,0.07619199752807618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,4,power_law_1.01,0.17422080039978027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,4,power_law_1.01,0.1700096011161804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,4,power_law_1.01,0.09281920194625855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,4,power_law_1.01,0.1736256003379822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,4,power_law_1.01,0.10343680381774903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,4,power_law_1.01,0.17253119945526124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,4,power_law_1.01,0.13199360370635987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,4,power_law_1.01,0.17868800163269044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,4,power_law_1.01,0.1313279986381531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,4,power_law_1.01,0.18271360397338868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,4,power_law_1.01,0.1308351993560791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,4,power_law_1.01,0.19288320541381837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,4,power_law_1.01,0.1386944055557251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,4,power_law_1.01,0.19607679843902587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,4,power_law_1.01,0.1365823984146118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,4,power_law_1.01,0.2044991970062256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,4,power_law_1.01,0.13873280286788942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,4,power_law_1.01,0.21656959056854247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,4,power_law_1.01,0.14257919788360596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,4,power_law_1.01,0.23701119422912598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,4,power_law_1.01,0.14268159866333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,4,power_law_1.01,0.2746495962142944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,4,power_law_1.01,0.1532863974571228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,4,power_law_1.01,0.30776960849761964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,4,power_law_1.01,0.1583680033683777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,4,power_law_1.01,0.3652992010116577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,4,power_law_1.01,0.43965439796447753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,4,power_law_1.01,0.5722623825073242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,4,power_law_1.01,0.698137617111206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,4,power_law_1.01,0.9833024024963379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,4,power_law_1.01,1.2854911804199218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,4,power_law_1.01,1.5232512474060058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,4,power_law_1.01,2.334329605102539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,4,power_law_1.01,4.807686233520508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,4,power_law_1.01,0.07209600210189819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,4,power_law_1.01,0.08753920197486878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,4,power_law_1.01,0.0741312026977539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,4,power_law_1.01,0.08613119721412658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,4,power_law_1.01,0.11034239530563354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,4,power_law_1.01,0.11917439699172974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,4,power_law_1.01,0.14848639965057372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,4,power_law_1.01,0.14702080488204955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,4,power_law_1.01,0.14759039878845215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,4,power_law_1.01,0.1474303960800171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,4,power_law_1.01,0.152019202709198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,4,power_law_1.01,0.16218240261077882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,4,power_law_1.01,0.16031999588012696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,4,power_law_1.01,0.16453759670257567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,4,power_law_1.01,0.16972800493240356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,4,power_law_1.01,0.1773568034172058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,4,power_law_1.01,0.1764224052429199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,4,power_law_1.01,0.1945088028907776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.01,0.2132863998413086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,4,power_law_1.01,0.021401600539684297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.01,0.24776320457458495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,4,power_law_1.01,0.024223999679088594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.01,0.2920192003250122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,4,power_law_1.01,0.025094398856163026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.01,0.3510272026062012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,4,power_law_1.01,0.030355200171470642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.01,0.4145472049713135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,4,power_law_1.01,0.03521279990673065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,4,power_law_1.01,0.04380159974098206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.01,0.5643648147583008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,4,power_law_1.01,0.04813440144062042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.01,0.7475711822509765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,4,power_law_1.01,0.052179199457168576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.01,0.9764224052429199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,4,power_law_1.01,0.05057920217514038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.01,1.357919979095459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,4,power_law_1.01,0.05338240265846252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,4,power_law_1.01,1.5498815536499024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,4,power_law_1.01,0.055820798873901366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,4,power_law_1.01,0.05648639798164368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,4,power_law_1.01,2.4823360443115234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,4,power_law_1.01,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,4,power_law_1.01,4.738291168212891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,4,power_law_1.01,0.06260480284690857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,4,power_law_1.01,0.07262719869613647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,4,power_law_1.01,0.07463039755821228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,4,power_law_1.01,0.07744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,4,power_law_1.01,0.11360640525817871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.01,0.11850240230560302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.01,0.13460479974746703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.01,0.1490880012512207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.01,0.1716223955154419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.01,0.19930880069732665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.01,0.25061759948730467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.01,0.30757761001586914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.01,0.4231935977935791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.01,0.540115213394165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,4,power_law_1.01,0.6533376216888428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,4,power_law_1.2,0.07901440262794494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,4,power_law_1.2,0.10051200389862061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,4,power_law_1.01,1.012441635131836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,4,power_law_1.2,0.0769919991493225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,4,power_law_1.01,1.9335039138793946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,4,power_law_1.2,0.0933568000793457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,4,power_law_1.2,0.10355199575424194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,4,power_law_1.2,0.11384320259094238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,4,power_law_1.2,0.15927040576934814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,4,power_law_1.2,0.16414719820022583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,4,power_law_1.2,0.16922240257263182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,4,power_law_1.01,0.16412800550460815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,4,power_law_1.2,0.16917760372161866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,4,power_law_1.01,0.18467199802398682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,4,power_law_1.2,0.16907520294189454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,4,power_law_1.01,0.20188798904418945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,4,power_law_1.2,0.17436800003051758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,4,power_law_1.01,0.24647040367126466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,4,power_law_1.2,0.18192000389099122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,4,power_law_1.01,0.2843839883804321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,4,power_law_1.2,0.18588160276412963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,4,power_law_1.01,0.3582655906677246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,4,power_law_1.2,0.1945024013519287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,4,power_law_1.01,0.44527359008789064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,4,power_law_1.2,0.19442559480667115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,4,power_law_1.01,0.5868544101715087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,4,power_law_1.2,0.20628480911254882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,4,power_law_1.01,0.7368192195892334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,4,power_law_1.2,0.22612481117248534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,4,power_law_1.01,1.045132827758789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,4,power_law_1.2,0.2397183895111084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,4,power_law_1.01,1.408409595489502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,4,power_law_1.2,0.29446399211883545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,4,power_law_1.01,1.7047615051269531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,4,power_law_1.2,0.32295680046081543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,4,power_law_1.01,2.6589311599731444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,4,power_law_1.2,0.3784192085266113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,4,power_law_1.2,0.44705920219421386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,4,power_law_1.01,5.201855850219727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,4,power_law_1.2,0.5975808143615723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,4,power_law_1.2,0.7693439960479737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,4,power_law_1.2,1.0667519569396973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,4,power_law_1.2,1.3527104377746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,4,power_law_1.2,1.8079744338989259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,4,power_law_1.2,2.7307647705078124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,4,power_law_1.2,5.2512256622314455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,4,power_law_1.2,0.06723200082778931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,4,power_law_1.2,0.08147839903831482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,4,power_law_1.2,0.06585599780082703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,4,power_law_1.2,0.07957119941711426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,4,power_law_1.2,0.08822399973869324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,4,power_law_1.2,0.1032639980316162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,4,power_law_1.2,0.12885760068893432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,4,power_law_1.2,0.13185919523239137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,4,power_law_1.2,0.12928639650344848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,4,power_law_1.2,0.13474559783935547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,4,power_law_1.2,0.13322880268096923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,4,power_law_1.2,0.1407807946205139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,4,power_law_1.2,0.1450816035270691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,4,power_law_1.2,0.14954880475997925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,4,power_law_1.2,0.15542399883270264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,4,power_law_1.2,0.15881600379943847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,4,power_law_1.2,0.17109760046005248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,4,power_law_1.2,0.07243520021438599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,4,power_law_1.2,0.18670079708099366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,4,power_law_1.2,0.21264638900756835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,4,power_law_1.2,0.08325120210647582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,4,power_law_1.2,0.2601855993270874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,4,power_law_1.2,0.06859520077705383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,4,power_law_1.2,0.29237120151519774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,4,power_law_1.2,0.0866752028465271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,4,power_law_1.2,0.1018496036529541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,4,power_law_1.2,0.3726655960083008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,4,power_law_1.2,0.11719679832458496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,4,power_law_1.2,0.46304001808166506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,4,power_law_1.2,0.14403200149536133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,4,power_law_1.2,0.6120448112487793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,4,power_law_1.2,0.1471295952796936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,4,power_law_1.2,0.8034496307373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,4,power_law_1.2,0.1471168041229248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,4,power_law_1.2,1.1223615646362304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,4,power_law_1.2,0.14985599517822265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,4,power_law_1.2,1.4677311897277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,4,power_law_1.2,0.15399680137634278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,4,power_law_1.2,1.8619199752807618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,4,power_law_1.2,0.16058239936828614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,4,power_law_1.2,2.6295679092407225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,4,power_law_1.2,0.16128000020980834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,4,power_law_1.2,0.16463359594345092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,4,power_law_1.2,6.0550590515136715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,4,power_law_1.2,0.17230080366134642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,4,power_law_1.2,0.17479039430618287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,4,power_law_1.2,0.18588160276412963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,4,power_law_1.2,0.20586879253387452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.2,0.222489595413208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.2,0.2563456058502197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.2,0.3048640012741089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.2,0.37007999420166016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.2,0.4461696147918701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,4,power_law_1.2,0.021849599480628968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.2,0.6549759864807129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,4,power_law_1.2,0.024691200256347655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.2,0.7208703994750977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,2,balanced,0.06671466430028279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,4,power_law_1.2,0.025158399343490602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,2,balanced,0.07216000060240428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.2,1.0515263557434082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,2,balanced,0.08310399949550629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,4,power_law_1.2,0.030905601382255555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.2,1.4828864097595216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,2,balanced,0.11117333173751831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,4,power_law_1.2,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,2,balanced,0.14618133505185446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,4,power_law_1.2,1.7312192916870117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,2,balanced,0.22894932826360068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,4,power_law_1.2,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,2,balanced,0.2208906610806783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,4,power_law_1.2,2.8188095092773438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,4,power_law_1.2,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,2,balanced,0.2241599957148234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,2,balanced,0.2226346731185913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,4,power_law_1.2,0.050886398553848265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,4,power_law_1.2,5.319295883178711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,2,balanced,0.222271998723348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,4,power_law_1.2,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,2,balanced,0.22186134258906046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,4,power_law_1.2,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,2,balanced,0.22565333048502603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,4,power_law_1.2,0.055283200740814206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,2,balanced,0.22597867250442505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,2,balanced,0.22910932699839273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,4,power_law_1.2,0.05839359760284424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,2,balanced,0.23981332778930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,4,power_law_1.2,0.0597055971622467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,2,balanced,0.2378986676534017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,4,power_law_1.2,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,4,power_law_1.2,0.07299839854240417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,2,balanced,0.24716800451278687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,4,power_law_1.2,0.07470719814300537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,2,balanced,0.2706986665725708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,4,power_law_1.2,0.07683839797973632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,2,balanced,0.28703467051188153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,4,power_law_1.2,0.11221120357513428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,2,balanced,0.32371199131011963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.2,0.14017280340194702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.2,0.1431167960166931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,2,balanced,0.35628799597422284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.2,0.14676480293273925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.2,0.1762943983078003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.2,0.19911680221557618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,2,balanced,0.43565332889556885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.2,0.2586303949356079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.2,0.31968638896942136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.2,0.42849278450012207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,2,balanced,0.5090399980545044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.2,0.5596096038818359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,4,power_law_1.2,0.6800127983093261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,4,power_law_1.2,1.0437376022338867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,2,balanced,0.7427466710408529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,4,power_law_1.2,2.048588752746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,2,balanced,0.8953920205434164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,2,balanced,1.2802399794260662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,2,balanced,1.5994240442911785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,2,balanced,1.963706652323405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,2,balanced,0.06019733349482218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,2,balanced,0.06357333560784657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,2,balanced,0.07132266461849213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,2,balanced,0.08851733803749084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,2,balanced,0.12379200259844463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,2,balanced,0.17485866943995157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,2,balanced,0.17575466632843018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,2,balanced,0.1759679913520813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,2,balanced,0.17997866868972778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,2,balanced,0.1807039976119995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,2,balanced,0.1815626621246338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,2,balanced,0.18522665898005167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,2,balanced,0.18433600664138794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,2,balanced,0.19133333365122476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,2,balanced,0.1981546680132548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,2,balanced,0.20467199881871542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,2,balanced,3.000101407368978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,2,balanced,0.21299733718236288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,2,balanced,0.23799467086791992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,2,balanced,0.07516799867153168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,2,balanced,0.2574080030123393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,2,balanced,0.08027733365694682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,2,balanced,0.0932426651318868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,2,balanced,0.3007199962933858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,2,balanced,0.11608533064524333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,2,balanced,0.17800533771514893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,2,balanced,0.2587786714235942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,2,balanced,0.2630400061607361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,2,balanced,0.3394506772359212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,2,balanced,0.2510346571604411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,2,balanced,0.252890666325887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,2,balanced,0.2603360017140706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,2,balanced,0.2504319945971171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,2,balanced,0.4305280049641927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,2,balanced,0.254037340482076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,2,balanced,0.2511199911435445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,2,balanced,0.2560853362083435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,2,balanced,0.262554665406545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,2,balanced,0.5165973504384359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,2,balanced,0.2589706579844157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,2,balanced,0.26714666684468585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,2,balanced,0.2906986673672994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,2,balanced,0.30671467383702594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,2,balanced,0.7521706422170004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,2,balanced,0.3519359827041626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,2,balanced,0.38628800710042316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,2,balanced,0.4732106526692708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,2,balanced,0.9247840245564779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,2,balanced,0.023599999646345775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,2,balanced,0.5569173494974772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,2,balanced,0.02495466669400533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,2,balanced,0.028586665789286297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,2,balanced,0.03224000086386999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,2,balanced,0.045226668318112694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,2,balanced,0.06656533479690552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,2,balanced,5.852501551310222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,2,balanced,0.06784533460934956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,2,balanced,0.8236160278320312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,2,balanced,0.07241599758466084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,2,balanced,0.07446399827798207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,2,balanced,0.0757173349459966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,2,balanced,1.3466453552246094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,2,balanced,0.08134933312733968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,2,balanced,0.08040533463160197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,2,balanced,0.08177066842714946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,2,balanced,0.9975840250651041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,2,balanced,0.08641599615414937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,2,balanced,0.08761066198348999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,2,balanced,0.09148800373077393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,2,balanced,0.09801066915194194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,2,balanced,0.12385599811871846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,2,balanced,0.13473066687583923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,2,balanced,1.4198880195617676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,2,balanced,0.18646933635075888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,2,balanced,1.733290672302246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,2,balanced,0.19973333676656088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,2,balanced,0.2544853289922078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,2,balanced,0.2876693407694499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,2,balanced,1.8511306444803874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,2,balanced,0.3301493326822917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,2,balanced,2.1311519940694175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,2,balanced,0.3749013344446818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,2,balanced,2.2749973932902017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,2,balanced,0.5951573451360067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,2,balanced,0.6953279972076416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,2,balanced,3.332592010498047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,2,balanced,3.5642665227254233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,2,balanced,0.9140799840291342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,2,power_law_1.01,0.07242239713668823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,2,power_law_1.01,0.088128000497818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,2,power_law_1.01,0.08142719864845276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,2,balanced,1.3771573702494304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,2,power_law_1.01,0.10308480262756348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,2,power_law_1.01,0.11958400011062623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,2,power_law_1.01,0.13972480297088624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,2,power_law_1.01,0.1902143955230713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,2,power_law_1.01,0.19298559427261353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,2,power_law_1.01,0.19189120531082154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,2,power_law_1.01,0.19215999841690062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,2,power_law_1.01,0.19692800045013428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,2,power_law_1.01,0.2080319881439209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,2,power_law_1.01,0.21196799278259276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,2,balanced,7.006079991658528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,2,power_law_1.01,0.21958398818969727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,2,balanced,6.548101425170898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,2,power_law_1.01,0.22811520099639893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,2,power_law_1.01,0.23243520259857178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,2,power_law_1.01,0.23969919681549073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,2,power_law_1.01,0.262719988822937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,2,power_law_1.01,0.2875648021697998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,2,power_law_1.01,0.3361216068267822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,2,power_law_1.01,0.3944767951965332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,2,power_law_1.01,0.4831232070922852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,2,power_law_1.01,0.5711487770080567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,2,power_law_1.01,0.7533120155334473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,2,power_law_1.01,0.9509823799133301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,2,power_law_1.01,1.3452927589416503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,2,power_law_1.01,1.7265344619750977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,2,power_law_1.01,2.0951168060302736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,2,power_law_1.01,3.353792190551758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,2,power_law_1.01,6.433030700683593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,2,balanced,2.7113866806030273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,2,power_law_1.01,0.0655232012271881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,2,power_law_1.01,0.07157760262489318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,2,power_law_1.01,0.070169597864151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,2,power_law_1.01,0.07998080253601074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,2,power_law_1.01,0.08661760091781616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,2,power_law_1.01,0.09207680225372314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,2,power_law_1.01,0.10328320264816285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,2,power_law_1.01,0.0850816011428833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,2,power_law_1.01,0.12105599641799927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,2,power_law_1.01,0.1087488055229187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,2,power_law_1.01,0.14720640182495118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,2,power_law_1.01,0.13670400381088257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,2,power_law_1.01,0.1526975989341736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,2,power_law_1.01,0.1677248001098633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,2,power_law_1.01,0.16000640392303467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,2,power_law_1.01,0.20182399749755858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,2,power_law_1.01,0.16062719821929933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,2,power_law_1.01,0.20667519569396972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,2,power_law_1.01,0.16416000127792357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,2,power_law_1.01,0.20965759754180907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,2,power_law_1.01,0.1718016028404236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,2,power_law_1.01,0.21328001022338866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,2,power_law_1.01,0.17293440103530883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,2,power_law_1.01,0.21270399093627929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,2,power_law_1.01,0.17844480276107788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,2,power_law_1.01,0.22428159713745116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,2,power_law_1.01,0.18738559484481812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,2,power_law_1.01,0.22499840259552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,2,power_law_1.01,0.19235199689865112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,2,power_law_1.01,0.2321984052658081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,2,power_law_1.01,0.20798718929290771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,2,power_law_1.01,0.2339263916015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,2,power_law_1.01,0.24510080814361573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,2,power_law_1.01,0.25603840351104734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,2,power_law_1.01,0.2840512037277222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.01,0.314630389213562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.01,0.36683518886566163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.01,0.42407679557800293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.01,0.5357312202453614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.01,0.6368832111358642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,2,power_law_1.2,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.01,0.8872768402099609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,2,power_law_1.2,0.08469120264053345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,2,power_law_1.01,0.021376000344753267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.01,1.1005824089050293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,2,power_law_1.2,0.08181759715080261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,2,power_law_1.01,0.02356480062007904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.01,1.5356608390808106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,2,power_law_1.2,0.09527040123939515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,2,power_law_1.01,0.02512640058994293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.01,2.0307519912719725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,2,power_law_1.2,0.1162559986114502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,2,power_law_1.01,0.030156800150871278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,2,power_law_1.01,2.487833595275879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,2,power_law_1.2,0.1371392011642456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,2,power_law_1.01,0.040479999780654904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,2,power_law_1.01,3.8261440277099608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,2,power_law_1.2,0.18079999685287476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,2,power_law_1.01,0.04560000002384186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,2,power_law_1.2,0.18685439825057984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,2,power_law_1.01,0.05671039819717407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,2,power_law_1.01,7.7940925598144535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,2,power_law_1.2,0.1903040051460266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,2,power_law_1.01,0.060262399911880496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,2,power_law_1.2,0.1958143949508667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,2,power_law_1.01,0.06221439838409424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,2,power_law_1.2,0.19955840110778808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,2,power_law_1.01,0.06445440053939819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,2,power_law_1.2,0.20856959819793702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,2,power_law_1.01,0.0705407977104187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,2,power_law_1.2,0.21226239204406738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,2,power_law_1.01,0.06810879707336426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,2,power_law_1.2,0.2185920000076294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,2,power_law_1.01,0.0686079978942871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,2,power_law_1.2,0.22958719730377197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,2,power_law_1.01,0.07779200077056884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,2,power_law_1.2,0.23291521072387694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,2,power_law_1.01,0.07442560195922851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,2,power_law_1.2,0.24661118984222413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,2,power_law_1.01,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,2,power_law_1.01,0.08610560297966004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,2,power_law_1.2,0.2665152072906494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,2,power_law_1.01,0.10321279764175414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,2,power_law_1.2,0.29087998867034914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.01,0.11498880386352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,2,power_law_1.2,0.3419327974319458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.01,0.17709439992904663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,2,power_law_1.2,0.40892801284790037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.01,0.1899839997291565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,2,power_law_1.2,0.4827263832092285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.01,0.20957438945770263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,2,power_law_1.2,0.624889612197876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,2,power_law_1.2,0.7989120006561279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.01,0.2378175973892212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.01,0.2890624046325684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,2,power_law_1.2,0.9524671554565429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.01,0.36480638980865476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,2,power_law_1.2,1.4278271675109864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.01,0.49248638153076174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,2,power_law_1.2,1.7890752792358398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.01,0.6231552124023437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,2,power_law_1.2,2.158835220336914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,2,power_law_1.01,0.7585855960845947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,2,power_law_1.2,3.1671104431152344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,2,power_law_1.01,1.1419391632080078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,2,power_law_1.2,7.2372993469238285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,2,power_law_1.01,2.237280082702637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,2,power_law_1.01,0.23172481060028077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,2,power_law_1.01,0.2593535900115967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,2,power_law_1.01,0.3135807991027832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,2,power_law_1.01,0.3731391906738281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,2,power_law_1.01,0.471999979019165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,2,power_law_1.01,0.59651198387146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,2,power_law_1.01,0.7811264038085938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,2,power_law_1.2,0.0639743983745575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,2,power_law_1.01,0.9982912063598632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,2,power_law_1.2,0.07277439832687378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,2,power_law_1.01,1.4369855880737306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,2,power_law_1.2,0.06934400200843811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,2,power_law_1.01,1.8268224716186523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,2,power_law_1.2,0.08279039859771728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,2,power_law_1.01,2.291263961791992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,2,power_law_1.2,0.09863680005073547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,2,power_law_1.01,3.5724159240722657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,2,power_law_1.2,0.11447679996490479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,2,power_law_1.2,0.14482560157775878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,2,power_law_1.01,6.945990753173828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,2,power_law_1.2,0.15294719934463502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,2,power_law_1.2,0.1545215964317322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,2,power_law_1.2,0.157043194770813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,2,power_law_1.2,0.1660223960876465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,2,power_law_1.2,0.1704767942428589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,2,power_law_1.2,0.17597440481185914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,2,power_law_1.2,0.18157440423965454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,2,power_law_1.2,0.18856960535049438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,2,power_law_1.2,0.19760639667510987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,2,power_law_1.2,0.2084223985671997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,2,power_law_1.2,0.2334912061691284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,2,power_law_1.2,0.021216000616550445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,2,power_law_1.2,0.2683327913284302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,2,power_law_1.2,0.022707200050354003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,2,power_law_1.2,0.025382399559020996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,2,power_law_1.2,0.33115520477294924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,2,power_law_1.2,0.3760384082794189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,2,power_law_1.2,0.03059839904308319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,2,power_law_1.2,0.07823359966278076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,2,power_law_1.2,0.0398719996213913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,2,power_law_1.2,0.5056191921234131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,2,power_law_1.2,0.09136639833450318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,2,power_law_1.2,0.044998401403427125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,2,power_law_1.2,0.6060671806335449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,2,power_law_1.2,0.08467199802398681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,2,power_law_1.2,0.05494400262832642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,2,power_law_1.2,0.7887680053710937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,2,power_law_1.2,0.058924800157546996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,2,power_law_1.2,0.10074880123138427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,2,power_law_1.2,0.9827775955200195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,2,power_law_1.2,0.0615231990814209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,2,power_law_1.2,0.1336127996444702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,2,power_law_1.2,1.4461824417114257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,2,power_law_1.2,0.06376320123672485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,2,power_law_1.2,0.1578112006187439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,2,power_law_1.2,1.893516731262207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,2,power_law_1.2,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,2,power_law_1.2,0.19833600521087646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,2,power_law_1.2,0.06716160178184509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,2,power_law_1.2,0.20538880825042724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,2,power_law_1.2,2.2695743560791017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,2,power_law_1.2,0.06888960003852844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,2,power_law_1.2,0.2124864101409912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,2,power_law_1.2,3.6426048278808594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,2,power_law_1.2,0.07850880026817322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,2,power_law_1.2,0.20983679294586183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,2,power_law_1.2,0.07586560249328614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,2,power_law_1.2,7.229138946533203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,2,power_law_1.2,0.21722879409790039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,2,power_law_1.2,0.07955200076103211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,2,power_law_1.2,0.22700800895690917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,2,power_law_1.2,0.08911359906196595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,2,power_law_1.2,0.2309567928314209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,2,power_law_1.2,0.1085055947303772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,2,power_law_1.2,0.23210880756378174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.2,0.12467199563980103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,2,power_law_1.2,0.24552319049835206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.2,0.18176000118255614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,2,power_law_1.2,0.24995839595794678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.2,0.19421440362930298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,2,power_law_1.2,0.2600640058517456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.2,0.21904640197753905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,2,power_law_1.2,0.29420158863067625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.2,0.32296319007873536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.2,0.2465087890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.2,0.38673279285430906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.2,0.3047424077987671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.2,0.43624320030212405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.2,0.36462080478668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.2,0.5611135959625244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.2,0.5074816226959229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.2,0.6981503963470459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.2,0.6317567825317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.2,0.8922816276550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,2,power_law_1.2,0.7990911960601806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.2,1.1003840446472168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,2,power_law_1.2,1.144371223449707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.2,1.687411117553711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,2,power_law_1.2,2.2053823471069336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.2,2.0770240783691407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,2,power_law_1.2,2.726617622375488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,2,power_law_1.2,3.8752254486083983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,2,power_law_1.2,8.108185577392579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,1,balanced,0.07910400132338206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,1,balanced,0.08363200227419536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,1,balanced,0.10328533252080281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,1,balanced,0.1343786617120107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,1,balanced,0.18078933159510294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,1,balanced,0.28458666801452637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,1,balanced,0.2856480081876119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,1,balanced,0.2836959958076477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,1,balanced,0.2890453338623047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,1,balanced,0.2897706627845764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,1,balanced,0.29257599512736004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,1,balanced,0.29793065786361694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,1,balanced,0.3006346623102824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,1,balanced,0.3060373266537984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,1,balanced,0.3158773382504781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,1,balanced,0.319269339243571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,1,balanced,0.332202672958374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,1,balanced,0.37193067868550617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,1,balanced,0.401253342628479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,1,balanced,0.4673440059026082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,1,balanced,0.5211999813715616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,1,balanced,0.6437280178070068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,1,balanced,0.7699946562449137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,1,balanced,0.07287466526031494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,1,balanced,0.07774400214354198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,1,balanced,0.09304533402125041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,1,balanced,0.12147200107574463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,1,balanced,1.1504639784495037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,1,balanced,0.16776533921559653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,1,balanced,0.2592693368593852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,1,balanced,0.2595946590105693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,1,balanced,0.26318933566411334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,1,balanced,0.26443199316660565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,1,balanced,0.2671626607577006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,1,balanced,0.2702560027440389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,1,balanced,0.27718400955200195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,1,balanced,1.4131360054016113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,1,balanced,0.2832213242848714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,1,balanced,0.28595733642578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,1,balanced,0.2956373294194539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,1,balanced,0.30110933383305866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,1,balanced,0.3131893277168274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,1,balanced,0.3568426767985026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,1,balanced,0.38837865988413495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,1,power_law_1.01,0.07198079824447631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,1,balanced,2.0280319849650064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,1,power_law_1.01,0.07832959890365601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,1,balanced,0.4604640007019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,1,power_law_1.01,0.09034879803657532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,1,power_law_1.01,0.11934720277786255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,1,power_law_1.01,0.06629760265350342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,1,balanced,0.5227466821670532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,1,power_law_1.01,0.14232319593429565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,1,power_law_1.01,0.07340800166130065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,1,power_law_1.01,0.16626559495925902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,1,power_law_1.01,0.08126720190048217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,1,power_law_1.01,0.22570879459381105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,1,power_law_1.01,0.11100800037384033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,1,balanced,0.663487990697225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,1,power_law_1.01,0.2327807903289795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,1,power_law_1.01,0.13293440341949464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,1,power_law_1.01,0.23631999492645264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,1,power_law_1.01,0.1612223982810974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,1,power_law_1.01,0.24705278873443604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,1,power_law_1.01,0.20353279113769532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,1,balanced,0.7977813084920248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,1,power_law_1.01,0.2575232028961182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,1,power_law_1.01,0.21500160694122314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,1,balanced,2.6265974044799805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,1,power_law_1.01,0.2611392021179199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,1,power_law_1.01,0.2212224006652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,1,power_law_1.01,0.26929919719696044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,1,power_law_1.01,0.22415359020233155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,1,power_law_1.01,0.27866239547729493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,1,power_law_1.01,0.2359999895095825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,1,balanced,1.2050080299377441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,1,power_law_1.01,0.28501760959625244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,1,power_law_1.01,0.24773120880126953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,1,power_law_1.01,0.24990079402923585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,1,power_law_1.01,0.29770240783691404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,1,power_law_1.01,0.25418241024017335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,1,power_law_1.01,0.3115520000457764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,1,power_law_1.01,0.26812798976898194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,1,power_law_1.01,0.35269761085510254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,1,power_law_1.01,0.27470719814300537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,1,power_law_1.01,0.3911360025405884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,1,balanced,1.4793386459350586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,1,power_law_1.01,0.29268479347229004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,1,power_law_1.01,0.47072639465332033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,1,power_law_1.01,0.339520001411438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,1,power_law_1.01,0.5587711811065674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,1,power_law_1.01,0.37829759120941164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,1,power_law_1.01,0.6982783794403076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,1,power_law_1.01,0.46554880142211913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,1,power_law_1.01,0.8352512359619141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,1,balanced,3.219327926635742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,1,power_law_1.01,0.5469312191009521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,1,power_law_1.01,1.1276608467102052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,1,power_law_1.01,0.7141952037811279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,1,power_law_1.01,1.4295680046081543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,1,power_law_1.01,0.8747903823852539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,1,balanced,2.158170700073242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,1,power_law_1.01,2.035622406005859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,1,power_law_1.01,1.2024895668029785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,1,power_law_1.01,2.616339111328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,1,power_law_1.01,1.5247232437133789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,1,power_law_1.01,3.2213184356689455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,1,power_law_1.01,2.1770496368408203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,1,power_law_1.01,5.002329635620117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,1,power_law_1.01,2.829580879211426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,1,power_law_1.01,9.81454086303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,1,power_law_1.01,3.4930240631103517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,1,power_law_1.01,5.468019104003906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,1,balanced,2.814938545227051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,1,power_law_1.01,10.765420532226562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,1,balanced,5.02345593770345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,1,balanced,3.4825121561686196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,1,balanced,5.46449089050293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,1,balanced,9.840426762898764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,1,balanced,10.749818166097006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,1,power_law_1.2,0.07083520293235779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,1,power_law_1.2,0.07904000282287597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,1,power_law_1.2,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,1,power_law_1.2,0.1152959942817688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,1,power_law_1.2,0.12974079847335815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,1,power_law_1.2,0.15794559717178344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,1,power_law_1.2,0.2208575963973999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,1,power_law_1.2,0.22930560111999512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,1,power_law_1.2,0.24030721187591553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,1,power_law_1.2,0.2479680061340332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,1,power_law_1.2,0.06618239879608154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,1,power_law_1.2,0.25169920921325684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,1,power_law_1.2,0.26220800876617434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,1,power_law_1.2,0.07251840233802795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,1,power_law_1.2,0.2699455976486206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,1,power_law_1.2,0.08252800107002259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,1,power_law_1.2,0.2781375885009766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,1,power_law_1.2,0.10355839729309083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,1,power_law_1.2,0.2906752109527588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,1,power_law_1.2,0.1251263976097107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,1,power_law_1.2,0.3007551908493042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,1,power_law_1.2,0.14852479696273804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,1,power_law_1.2,0.31731839179992677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,1,power_law_1.2,0.2001471996307373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,1,power_law_1.2,0.36106879711151124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,1,power_law_1.2,0.21246719360351562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,1,power_law_1.2,0.39825921058654784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,1,power_law_1.2,0.22074880599975585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,1,power_law_1.2,0.4924928188323975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,1,power_law_1.2,0.22614400386810302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,1,power_law_1.2,0.5665023803710938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,1,power_law_1.2,0.23331201076507568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,1,power_law_1.2,0.7059840202331543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,1,power_law_1.2,0.24759039878845215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,1,power_law_1.2,0.25167360305786135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,1,power_law_1.2,0.8551103591918945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,1,power_law_1.2,0.2614720106124878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,1,power_law_1.2,1.1397120475769043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,1,power_law_1.2,0.27146880626678466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,1,power_law_1.2,1.437446403503418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,1,power_law_1.2,0.28349440097808837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,1,power_law_1.2,2.0372800827026367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,1,power_law_1.2,0.3026560068130493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,1,power_law_1.2,2.635577583312988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,1,power_law_1.2,0.35281920433044434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,1,power_law_1.2,3.2147518157958985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,1,power_law_1.2,0.389299201965332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,1,power_law_1.2,5.0044097900390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,1,power_law_1.2,0.4778048038482666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,1,power_law_1.2,0.5569087982177734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,1,power_law_1.2,9.810348510742188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,1,power_law_1.2,0.7202367782592773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,1,power_law_1.2,0.8888192176818848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,1,power_law_1.2,1.207692813873291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,1,power_law_1.2,1.5314175605773925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,1,power_law_1.2,2.179609680175781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,1,power_law_1.2,2.8263744354248046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,1,power_law_1.2,3.482342529296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,1,power_law_1.2,5.458412933349609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,1,power_law_1.2,10.769920349121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,32,balanced,0.03403733422358831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,32,balanced,0.034346667428811394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,32,balanced,0.03440533330043157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,32,balanced,0.03868799904982249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,32,balanced,0.04451199869314829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,32,balanced,0.043136000633239746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,32,balanced,0.04094400008519491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,32,balanced,0.04269866645336151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,32,balanced,0.04251199960708618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,32,balanced,0.04445866743723551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,32,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,32,balanced,0.0444213350613912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,32,balanced,0.04238399863243103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,32,balanced,0.044266665975252785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,32,balanced,0.04649066428343455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,32,balanced,0.04442666471004486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,32,balanced,0.046469335754712425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,32,balanced,0.046341334780057274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,32,balanced,0.049029335379600525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,32,balanced,0.04665066798528036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,32,balanced,0.04870399832725525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,32,balanced,0.05880533158779144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,32,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,32,balanced,0.058837334314982094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,32,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,32,balanced,0.03409066547950109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,32,balanced,0.038890667259693146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,32,balanced,0.038480001191298165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,32,balanced,0.04061333338419596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,32,balanced,0.04233600199222565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,32,balanced,0.04074133435885111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,32,balanced,0.07353066901365916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,32,balanced,0.040634666879971824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,32,balanced,0.040549332896868386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,32,balanced,0.042410666743914284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,32,balanced,0.042634665966033936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,32,balanced,0.04125866790612539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,32,balanced,0.04268800218900045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,32,balanced,0.046336000164349876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,32,balanced,0.08965333302815755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,32,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,32,balanced,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,32,balanced,0.046682665745417275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,32,balanced,0.05023466547330221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,32,balanced,0.05115733544031779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,32,balanced,0.11170132954915364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,32,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,32,balanced,0.06271466612815857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,32,balanced,0.07257600128650665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,32,balanced,0.13677866260210672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,32,balanced,0.09918399651845296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,32,balanced,0.12414933244387309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,32,balanced,0.033861334125200905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,32,balanced,0.034661332766215004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,32,balanced,0.03455466777086258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,32,balanced,0.036848001182079315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,32,balanced,0.04109866668780645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,32,balanced,0.04225599765777588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,32,balanced,0.04144000013669332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,32,balanced,0.04413333535194397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,32,balanced,0.04307733476161957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,32,balanced,0.04087999959786733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,32,balanced,0.16291200121243796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,32,balanced,0.16249066591262817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,32,balanced,0.04267199834187826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,32,balanced,0.04229333500067393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,32,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,32,balanced,0.042538667718569435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,32,balanced,0.04640533526738485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,32,balanced,0.045738667249679565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,32,balanced,0.0462773342927297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,32,balanced,0.04845866560935974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,32,balanced,0.048063998421033226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,32,balanced,0.2060799996058146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,32,balanced,0.019391999890406925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,32,balanced,0.022154666483402252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,32,balanced,0.052485331892967224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,32,balanced,0.020330666253964107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,32,balanced,0.020319999506076176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,32,balanced,0.054197331269582115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,32,balanced,0.021744000415007275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,32,balanced,0.024122667809327442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,32,balanced,0.03477333237727483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,32,balanced,0.03433600068092346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,32,balanced,0.0642080008983612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,32,balanced,0.03452266752719879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,32,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,32,balanced,0.033743999898433685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,32,balanced,0.03177600105603536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,32,balanced,0.068271999557813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,32,balanced,0.03049066662788391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,32,balanced,0.2288586695988973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,32,balanced,0.030373332401116688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,32,balanced,0.2500159939130147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,32,balanced,0.03681066632270813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,32,balanced,0.0363520011305809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,32,balanced,0.08713600039482117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,32,balanced,0.034346667428811394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,32,balanced,0.038245332737763725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,32,balanced,0.04645866652329763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,32,balanced,0.1037493348121643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,32,balanced,0.050245334704717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,32,balanced,0.05892266829808553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,32,balanced,0.07316266496976216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,32,balanced,0.1357493301232656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,32,balanced,0.09300266702969869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,32,balanced,0.3665119806925456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,32,balanced,0.12242133418718974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,32,balanced,0.164383997519811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,32,balanced,0.15405866503715515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,32,balanced,0.2161173423131307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,32,balanced,0.19157866636912027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,32,balanced,0.41659732659657794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,32,balanced,0.2775733272234599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,32,balanced,0.2795413335164388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,32,balanced,0.6884480317433676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,32,balanced,0.3417439858118693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,32,balanced,0.5309173266092936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,32,balanced,0.5120640198389689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,32,balanced,1.0357120037078857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,32,power_law_1.01,0.03709439933300018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,32,power_law_1.01,0.03596160113811493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,32,power_law_1.01,0.034297600388526917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,32,power_law_1.01,0.036550399661064145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,32,power_law_1.01,0.03790720105171204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,32,power_law_1.01,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,32,power_law_1.01,0.04021120071411133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,32,power_law_1.01,0.0409280002117157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,32,power_law_1.01,0.04110719859600067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,32,power_law_1.01,0.04168320000171662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,32,power_law_1.01,0.04049279987812042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,32,power_law_1.01,0.041152000427246094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,32,power_law_1.01,0.041459199786186215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,32,power_law_1.01,0.03829759955406189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,32,power_law_1.01,0.04259200096130371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,32,power_law_1.01,0.035488000512123107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,32,power_law_1.01,0.045100799202919005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,32,power_law_1.01,0.03258239924907684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,32,power_law_1.01,0.04500479996204376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,32,power_law_1.01,0.035897600650787356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,32,power_law_1.01,0.04679040014743805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,32,power_law_1.01,0.03697279989719391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,32,power_law_1.01,0.04904319941997528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,32,power_law_1.01,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,32,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,32,power_law_1.01,0.05358719825744629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,32,power_law_1.01,0.03763200044631958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,32,power_law_1.01,0.0593280017375946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,32,power_law_1.01,0.038726401329040525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,32,power_law_1.01,0.06867200136184692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,32,power_law_1.01,0.039699199795722964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,32,power_law_1.01,0.08225280046463013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,32,power_law_1.01,0.040403199195861814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,32,power_law_1.01,0.09901440143585205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,32,power_law_1.01,0.040678399801254275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,32,power_law_1.01,0.11839359998703003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,32,power_law_1.01,0.040601599216461184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,32,power_law_1.01,0.14039039611816406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,32,power_law_1.01,0.04217599928379059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,32,power_law_1.01,0.21149439811706544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,32,power_law_1.01,0.04444800019264221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,32,power_law_1.01,0.2581183910369873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,32,power_law_1.01,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,32,power_law_1.01,0.32389121055603026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,32,power_law_1.01,0.047679999470710756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,32,power_law_1.01,0.4799680233001709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,32,power_law_1.01,0.05055360198020935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,32,power_law_1.01,0.054048001766204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,32,power_law_1.01,0.9844351768493652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,32,power_law_1.01,0.06199679970741272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,32,power_law_1.01,0.06958720088005066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,32,power_law_1.01,0.09249280095100403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,32,power_law_1.01,0.11025279760360718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,32,power_law_1.01,0.14746240377426148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,32,power_law_1.01,0.1975167989730835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,32,power_law_1.01,0.2637120008468628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,32,power_law_1.01,0.31562879085540774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,32,power_law_1.01,0.4055935859680176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,32,power_law_1.01,0.6435776233673096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,32,power_law_1.01,1.0881279945373534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,32,power_law_1.01,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,32,power_law_1.01,0.03744640052318573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,32,power_law_1.01,0.03463680148124695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,32,power_law_1.01,0.036735999584198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,32,power_law_1.01,0.039027199149131775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,32,power_law_1.01,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,32,power_law_1.01,0.0395583987236023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,32,power_law_1.01,0.040678399801254275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,32,power_law_1.01,0.040140798687934874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,32,power_law_1.01,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,32,power_law_1.01,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,32,power_law_1.01,0.04115839898586273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,32,power_law_1.01,0.040870401263237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,32,power_law_1.01,0.042822399735450746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,32,power_law_1.01,0.04503040015697479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,32,power_law_1.01,0.04643200039863586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,32,power_law_1.01,0.04814079999923706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,32,power_law_1.01,0.052537602186203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,32,power_law_1.01,0.05537279844284058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,32,power_law_1.01,0.06330879926681518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,32,power_law_1.01,0.07490559816360473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,32,power_law_1.01,0.08736640214920044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,32,power_law_1.01,0.0979968011379242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,32,power_law_1.01,0.1291264057159424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,32,power_law_1.01,0.14893440008163453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,32,power_law_1.01,0.20377600193023682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,32,power_law_1.01,0.26054399013519286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,32,power_law_1.01,0.017215999960899352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,32,power_law_1.01,0.32233600616455077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,32,power_law_1.01,0.016358399391174318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,32,power_law_1.01,0.018764799833297728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,32,power_law_1.01,0.5665215969085693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,32,power_law_1.01,0.018892799317836762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,32,power_law_1.01,0.9583552360534668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,32,power_law_1.01,0.01886080056428909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,32,power_law_1.01,0.022655999660491942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,32,power_law_1.01,0.03296639919281006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,32,power_law_1.01,0.03237119913101196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,32,power_law_1.01,0.03163520097732544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,32,power_law_1.01,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,32,power_law_1.01,0.03312639892101288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,32,power_law_1.01,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,32,power_law_1.01,0.0324288010597229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,32,power_law_1.01,0.030707201361656188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,32,power_law_1.01,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,32,power_law_1.01,0.03287039995193482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,32,power_law_1.01,0.03313280045986176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,32,power_law_1.01,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,32,power_law_1.01,0.0364544004201889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,32,power_law_1.01,0.0460671991109848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,32,power_law_1.01,0.04890879988670349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,32,power_law_1.01,0.05777279734611511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,32,power_law_1.01,0.07125120162963867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,32,power_law_1.01,0.09423999786376953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,32,power_law_1.01,0.11537920236587525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,32,power_law_1.01,0.15626239776611328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,32,power_law_1.01,0.20409600734710692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,32,power_law_1.01,0.24401919841766356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,32,power_law_1.01,0.38346879482269286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,32,power_law_1.01,0.7640063762664795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,32,power_law_1.2,0.04121600091457367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,32,power_law_1.2,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,32,power_law_1.2,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,32,power_law_1.2,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,32,power_law_1.2,0.0374208003282547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,32,power_law_1.2,0.04116480052471161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,32,power_law_1.2,0.041068801283836366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,32,power_law_1.2,0.040268799662590025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,32,power_law_1.2,0.04099839925765991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,32,power_law_1.2,0.04097279906272888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,32,power_law_1.2,0.04192639887332916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,32,power_law_1.2,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,32,power_law_1.2,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,32,power_law_1.2,0.042905598878860474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,32,power_law_1.2,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,32,power_law_1.2,0.04508799910545349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,32,power_law_1.2,0.04849919974803925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,32,power_law_1.2,0.05495679974555969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,32,power_law_1.2,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,32,power_law_1.2,0.05589759945869446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,32,power_law_1.2,0.032332798838615416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,32,power_law_1.2,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,32,power_law_1.2,0.033395200967788696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,32,power_law_1.2,0.08090879917144775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,32,power_law_1.2,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,32,power_law_1.2,0.0948095977306366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,32,power_law_1.2,0.035097599029541016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,32,power_law_1.2,0.10613759756088256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,32,power_law_1.2,0.03870719969272614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,32,power_law_1.2,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,32,power_law_1.2,0.14302719831466676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,32,power_law_1.2,0.03909119963645935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,32,power_law_1.2,0.19646079540252687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,32,power_law_1.2,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,32,power_law_1.2,0.24588799476623535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,32,power_law_1.2,0.03971840143203735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,32,power_law_1.2,0.44341120719909666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,32,power_law_1.2,0.03992320001125336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,32,power_law_1.2,0.5315392017364502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,32,power_law_1.2,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,32,power_law_1.2,0.8223936080932617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,32,power_law_1.2,0.040659201145172116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,32,power_law_1.2,2.0004671096801756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,32,power_law_1.2,0.04128639996051788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,32,power_law_1.2,0.045075199007987975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,32,power_law_1.2,0.046540799736976626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,32,power_law_1.2,0.04785920083522797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,32,power_law_1.2,0.05303040146827698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,32,power_law_1.2,0.055718398094177245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,32,power_law_1.2,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,32,power_law_1.2,0.07597439885139465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,32,power_law_1.2,0.09500160217285156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,32,power_law_1.2,0.1312064051628113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,32,power_law_1.2,0.1901695966720581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,32,power_law_1.2,0.231660795211792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,32,power_law_1.2,0.3052031993865967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,32,power_law_1.2,0.49779839515686036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,32,power_law_1.2,0.568665599822998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,32,power_law_1.2,1.0496576309204102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,32,power_law_1.2,2.2719743728637694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,32,power_law_1.2,0.03944959938526153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,32,power_law_1.2,0.03266560137271881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,32,power_law_1.2,0.034297600388526917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,32,power_law_1.2,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,32,power_law_1.2,0.037459200620651244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,32,power_law_1.2,0.04046080112457275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,32,power_law_1.2,0.03994239866733551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,32,power_law_1.2,0.04074879884719849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,32,power_law_1.2,0.03969280123710632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,32,power_law_1.2,0.04063999950885773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,32,power_law_1.2,0.04154880046844482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,32,power_law_1.2,0.04248960018157959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,32,power_law_1.2,0.04166400134563446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,32,power_law_1.2,0.04230400025844574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,32,power_law_1.2,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,32,power_law_1.2,0.046963199973106384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,32,power_law_1.2,0.04986239969730377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,32,power_law_1.2,0.0549888014793396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,32,power_law_1.2,0.05860480070114136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,32,power_law_1.2,0.06496000289916992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,32,power_law_1.2,0.07593600153923034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,32,power_law_1.2,0.017529599368572235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,32,power_law_1.2,0.09319679737091065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,32,power_law_1.2,0.01687040030956268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,32,power_law_1.2,0.10782719850540161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,32,power_law_1.2,0.01791999936103821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,32,power_law_1.2,0.14654719829559326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,32,power_law_1.2,0.018265600502490997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,32,power_law_1.2,0.018828800320625304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,32,power_law_1.2,0.18565759658813477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,32,power_law_1.2,0.022412799298763275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,32,power_law_1.2,0.3143232107162476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,32,power_law_1.2,0.03314560055732727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,32,power_law_1.2,0.3611455917358398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,32,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,32,power_law_1.2,0.4212992191314697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,32,power_law_1.2,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,32,power_law_1.2,0.8098367691040039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,32,power_law_1.2,0.0321727991104126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,32,power_law_1.2,2.1778432846069338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,32,power_law_1.2,0.032313600182533264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,32,power_law_1.2,0.032307198643684386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,32,power_law_1.2,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,32,power_law_1.2,0.030585598945617676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,32,power_law_1.2,0.03185279965400696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,32,power_law_1.2,0.030777600407600404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,32,power_law_1.2,0.03238399922847748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,32,power_law_1.2,0.034720000624656674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,32,power_law_1.2,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,32,power_law_1.2,0.04794879853725433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,32,power_law_1.2,0.049804800748825075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,32,power_law_1.2,0.06108800172805786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,32,power_law_1.2,0.07212799787521362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,32,power_law_1.2,0.10195839405059814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,32,power_law_1.2,0.12471679449081421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,32,power_law_1.2,0.17932159900665284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,32,power_law_1.2,0.24949119091033936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,32,power_law_1.2,0.2748415946960449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,32,power_law_1.2,0.46720638275146487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,32,power_law_1.2,0.8813695907592773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,16,balanced,0.03218133250872294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,16,balanced,0.03313600023587545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,16,balanced,0.03533333291610082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,16,balanced,0.038746667404969536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,16,balanced,0.04244266450405121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,16,balanced,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,16,balanced,0.04260799785455068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,16,balanced,0.04252799848715464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,16,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,16,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,16,balanced,0.04313066601753235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,16,balanced,0.04286933441956838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,16,balanced,0.04461866617202759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,16,balanced,0.04394666850566864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,16,balanced,0.04513066510359446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,16,balanced,0.046485334634780884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,16,balanced,0.04598933458328247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,16,balanced,0.046480000019073486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,16,balanced,0.04655466477076212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,16,balanced,0.05035200218359629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,16,balanced,0.03226666649182638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,16,balanced,0.032986665765444435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,16,balanced,0.03456533451875051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,16,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,16,balanced,0.036533333361148834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,16,balanced,0.04085333396991094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,16,balanced,0.040405333042144775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,16,balanced,0.04045866678158442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,16,balanced,0.062352001667022705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,16,balanced,0.041493333876132965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,16,balanced,0.040864000717798867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,16,balanced,0.04114133367935816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,16,balanced,0.040906667709350586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,16,balanced,0.06643199920654297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,16,balanced,0.040778666734695435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,16,balanced,0.0425546665986379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,16,balanced,0.04254400233427683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,16,balanced,0.04435733457406362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,16,balanced,0.08524266878763835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,16,balanced,0.047024001677831016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,16,balanced,0.04713066418965658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,16,balanced,0.04903466502825419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,16,balanced,0.04934399823347727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,16,balanced,0.10348266363143921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,16,balanced,0.055589333176612854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,16,balanced,0.058933332562446594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,16,balanced,0.0711946686108907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,16,balanced,0.12868266304334006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,16,balanced,0.07914666831493378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,16,balanced,0.11584533254305522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,16,balanced,0.15898133317629495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,16,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,16,balanced,0.034202667574087776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,16,balanced,0.034474665919939675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,16,balanced,0.1407786707083384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,16,balanced,0.03874133278926214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,16,balanced,0.04543999830881754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,16,balanced,0.04450133442878723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,16,balanced,0.04453866680463155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,16,balanced,0.045034666856129967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,16,balanced,0.044906665881474815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,16,balanced,0.04489600161711375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,16,balanced,0.044778664906819664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,16,balanced,0.18963199853897095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,16,balanced,0.04469866553942362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,16,balanced,0.04677333434422811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,16,balanced,0.04651733239491781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,16,balanced,0.048709332942962646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,16,balanced,0.190720001856486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,16,balanced,0.05020266771316528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,16,balanced,0.05038933455944061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,16,balanced,0.05235733091831207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,16,balanced,0.053029333551724754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,16,balanced,0.05862933397293091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,16,balanced,0.06112533311049143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,16,balanced,0.017674667139848072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,16,balanced,0.02027733375628789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,16,balanced,0.07392000158627827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,16,balanced,0.019610666980346043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,16,balanced,0.01995733380317688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,16,balanced,0.019632000476121902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,16,balanced,0.2388746738433838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,16,balanced,0.021898667017618816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,16,balanced,0.2797866662343343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,16,balanced,0.08045333127180736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,16,balanced,0.023589332898457844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,16,balanced,0.02387733260790507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,16,balanced,0.035349334279696144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,16,balanced,0.034586665530999504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,16,balanced,0.03604800005753835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,16,balanced,0.034186666210492454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,16,balanced,0.1074666678905487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,16,balanced,0.03262399882078171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,16,balanced,0.03289066751797994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,16,balanced,0.03804266701141993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,16,balanced,0.03730133424202601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,16,balanced,0.036288000643253326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,16,balanced,0.1270080010096232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,16,balanced,0.03862400104602178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,16,balanced,0.044293334086736046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,16,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,16,balanced,0.28780800104141235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,16,balanced,0.05253866811593374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,16,balanced,0.16754132509231567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,16,balanced,0.06861333549022675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,16,balanced,0.07913599908351898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,16,balanced,0.20841066042582193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,16,balanced,0.10897599657376607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,16,balanced,0.1321440041065216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,16,balanced,0.5168799956639608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,16,balanced,0.4291306734085083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,16,balanced,0.24836266040802002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,16,balanced,0.18477867046991983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,16,balanced,0.2367146611213684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,16,balanced,0.36740267276763916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,16,balanced,0.2868266701698303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,16,balanced,0.4432053168614705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,16,balanced,0.8329599698384603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,16,balanced,0.6928160190582275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,16,power_law_1.01,0.0403328001499176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,16,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,16,balanced,0.8572533130645752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,16,power_law_1.01,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,16,power_law_1.01,0.035776001214981076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,16,power_law_1.01,0.03709439933300018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,16,power_law_1.01,0.04054400026798248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,16,power_law_1.01,0.039724799990653994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,16,power_law_1.01,0.041068801283836366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,16,power_law_1.01,0.03964160084724426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,16,power_law_1.01,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,16,power_law_1.01,0.04257279932498932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,16,power_law_1.01,0.041388800740242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,16,power_law_1.01,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,16,power_law_1.01,0.042182400822639465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,16,power_law_1.01,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,16,power_law_1.01,0.04624640047550201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,16,power_law_1.01,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,16,power_law_1.01,0.050367999076843264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,16,power_law_1.01,0.05356159806251526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,16,power_law_1.01,0.0602944016456604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,16,power_law_1.01,0.06621440052986145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,16,power_law_1.01,0.0847104012966156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,16,power_law_1.01,0.09466879963874816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,16,power_law_1.01,0.12955520153045655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,16,power_law_1.01,0.14203519821166993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,16,power_law_1.01,0.21986560821533202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,16,power_law_1.01,0.2847615957260132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,16,power_law_1.01,0.2992703914642334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,16,power_law_1.01,0.49921278953552245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,16,power_law_1.01,1.055513572692871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,16,power_law_1.01,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,16,power_law_1.01,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,16,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,16,power_law_1.01,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,16,power_law_1.01,0.03686400055885315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,16,power_law_1.01,0.04284160137176514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,16,power_law_1.01,0.0387584000825882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,16,power_law_1.01,0.0385343998670578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,16,power_law_1.01,0.03895680010318756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,16,power_law_1.01,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,16,power_law_1.01,0.03946239948272705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,16,power_law_1.01,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,16,power_law_1.01,0.04017919898033142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,16,power_law_1.01,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,16,power_law_1.01,0.04101119935512543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,16,power_law_1.01,0.04099839925765991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,16,power_law_1.01,0.0416128009557724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,16,power_law_1.01,0.04188160002231598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,16,power_law_1.01,0.04614399969577789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,16,power_law_1.01,0.04764159917831421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,16,power_law_1.01,0.0486272007226944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,16,power_law_1.01,0.05317760109901428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,16,power_law_1.01,0.05675519704818725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,16,power_law_1.01,0.0667967975139618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,16,power_law_1.01,0.071424001455307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,16,power_law_1.01,0.08946560025215149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,16,power_law_1.01,0.11006720066070556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,16,power_law_1.01,0.15367679595947265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,16,power_law_1.01,0.20797441005706788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,16,power_law_1.01,0.27918078899383547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,16,power_law_1.01,0.3212224006652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,16,power_law_1.01,0.42986240386962893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,16,power_law_1.01,0.6967679977416992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,16,power_law_1.01,1.3744256019592285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,16,power_law_1.01,0.017203199863433837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,16,power_law_1.01,0.017375999689102174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,16,power_law_1.01,0.018374399840831758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,16,power_law_1.01,0.017632000148296356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,16,power_law_1.01,0.01892479956150055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,16,power_law_1.01,0.020339199900627138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,16,power_law_1.01,0.023104000091552734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,16,power_law_1.01,0.0219200000166893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,16,power_law_1.01,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,16,power_law_1.01,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,16,power_law_1.01,0.033907198905944826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,16,power_law_1.01,0.03324800133705139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,16,power_law_1.01,0.03249920010566711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,16,power_law_1.01,0.03191039860248566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,16,power_law_1.01,0.03160319924354553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,16,power_law_1.01,0.033657601475715636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,16,power_law_1.01,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,16,power_law_1.01,0.03612160086631775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,16,power_law_1.01,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,16,power_law_1.01,0.04203520119190216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,16,power_law_1.01,0.04561919867992401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,16,power_law_1.01,0.05731199979782105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,16,power_law_1.01,0.06467840075492859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,16,power_law_1.01,0.08604159951210022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,16,power_law_1.01,0.1030079960823059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,16,power_law_1.01,0.13980799913406372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,16,power_law_1.01,0.17902719974517822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,16,power_law_1.01,0.22343039512634277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,16,power_law_1.01,0.3337280035018921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,16,power_law_1.01,0.6483007907867432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,16,power_law_1.2,0.03971840143203735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,16,power_law_1.2,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,16,power_law_1.2,0.03553279936313629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,16,power_law_1.2,0.03654400110244751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,16,power_law_1.2,0.03685120046138764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,16,power_law_1.2,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,16,power_law_1.2,0.039680001139640805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,16,power_law_1.2,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,16,power_law_1.2,0.040428799390792844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,16,power_law_1.2,0.04197759926319122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,16,power_law_1.2,0.04206719994544983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,16,power_law_1.2,0.04172160029411316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,16,power_law_1.2,0.04154880046844482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,16,power_law_1.2,0.04261760115623474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,16,power_law_1.2,0.04578559994697571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,16,power_law_1.2,0.046412798762321475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,16,power_law_1.2,0.049619200825691226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,16,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,16,power_law_1.2,0.05459840297698974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,16,power_law_1.2,0.06168320178985596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,16,power_law_1.2,0.06592000126838685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,16,power_law_1.2,0.08566399812698364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,16,power_law_1.2,0.10257920026779174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,16,power_law_1.2,0.13301119804382325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,16,power_law_1.2,0.16403839588165284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,16,power_law_1.2,0.25921919345855715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,16,power_law_1.2,0.35859839916229247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,16,power_law_1.2,0.40247678756713867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,16,power_law_1.2,0.677836799621582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,16,power_law_1.2,1.7539264678955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,16,power_law_1.01,0.03943040072917938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,16,power_law_1.01,0.04330880045890808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,16,power_law_1.01,0.043968001008033754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,16,power_law_1.01,0.04293760061264038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,16,power_law_1.01,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,16,power_law_1.2,0.037171199917793274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,16,power_law_1.01,0.043296000361442565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,16,power_law_1.2,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,16,power_law_1.01,0.043558400869369504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,16,power_law_1.2,0.03264000117778778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,16,power_law_1.01,0.04475519955158234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,16,power_law_1.2,0.03498879969120026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,16,power_law_1.01,0.045535999536514285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,16,power_law_1.2,0.03646720051765442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,16,power_law_1.01,0.045286399126052854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,16,power_law_1.2,0.03829759955406189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,16,power_law_1.01,0.04844799935817719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,16,power_law_1.2,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,16,power_law_1.01,0.0497408002614975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,16,power_law_1.2,0.038726401329040525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,16,power_law_1.01,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,16,power_law_1.2,0.03953279852867127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,16,power_law_1.01,0.057158398628234866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,16,power_law_1.2,0.03928320109844208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,16,power_law_1.01,0.061990398168563846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,16,power_law_1.2,0.04073599874973297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,16,power_law_1.01,0.06961920261383056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,16,power_law_1.2,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,16,power_law_1.01,0.07793279886245727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,16,power_law_1.2,0.040454399585723874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,16,power_law_1.01,0.09548799991607666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,16,power_law_1.2,0.04106239974498749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,16,power_law_1.01,0.10977280139923096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,16,power_law_1.2,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,16,power_law_1.01,0.1412992000579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,16,power_law_1.2,0.047577598690986635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,16,power_law_1.01,0.1793984055519104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,16,power_law_1.2,0.04999679923057556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,16,power_law_1.01,0.23980801105499266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,16,power_law_1.2,0.05425919890403748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,16,power_law_1.01,0.3385855913162231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,16,power_law_1.2,0.0591808021068573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,16,power_law_1.01,0.4073728084564209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,16,power_law_1.2,0.06627839803695679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,16,power_law_1.01,0.7073599815368652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,16,power_law_1.2,0.07822719812393189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,16,power_law_1.2,0.0986624002456665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,16,power_law_1.01,1.244102382659912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,16,power_law_1.2,0.12283519506454468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,16,power_law_1.2,0.18256640434265137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,16,power_law_1.2,0.24844160079956054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,16,power_law_1.2,0.309881591796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,16,power_law_1.2,0.4758912086486816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,16,power_law_1.2,0.5690815925598145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,16,power_law_1.2,1.0138239860534668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,16,power_law_1.2,1.7484224319458008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,16,power_law_1.2,0.04134399890899658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,16,power_law_1.2,0.03489919900894165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,16,power_law_1.2,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,16,power_law_1.2,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,16,power_law_1.2,0.03841919898986816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,16,power_law_1.2,0.042368000745773314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,16,power_law_1.2,0.04347519874572754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,16,power_law_1.2,0.04154880046844482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,16,power_law_1.2,0.044588801264762876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,16,power_law_1.2,0.04288640022277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,16,power_law_1.2,0.04490880072116852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,16,power_law_1.2,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,16,power_law_1.2,0.04498560130596161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,16,power_law_1.2,0.0455487996339798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,16,power_law_1.2,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,16,power_law_1.2,0.05072000026702881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,16,power_law_1.2,0.053011202812194826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,16,power_law_1.2,0.05953279733657837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,16,power_law_1.2,0.06225280165672302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,16,power_law_1.2,0.07098879814147949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,16,power_law_1.2,0.07904639840126038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,16,power_law_1.2,0.10238080024719239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,16,power_law_1.2,0.1302783966064453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,16,power_law_1.2,0.1754431962966919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,16,power_law_1.2,0.23074560165405272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,16,power_law_1.2,0.30441598892211913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,16,power_law_1.2,0.5048768043518066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,16,power_law_1.2,0.45722241401672364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,16,power_law_1.2,1.0675968170166015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,16,power_law_1.2,1.9505535125732423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,16,power_law_1.2,0.01788160055875778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,16,power_law_1.2,0.018195199966430663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,16,power_law_1.2,0.017836800217628478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,16,power_law_1.2,0.018745599687099455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,16,power_law_1.2,0.021132799983024596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,16,power_law_1.2,0.020793600380420683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,16,power_law_1.2,0.023558400571346283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,16,power_law_1.2,0.023039999604225158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,16,power_law_1.2,0.034195199608802795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,16,power_law_1.2,0.03384959995746613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,16,power_law_1.2,0.03492479920387268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,16,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,16,power_law_1.2,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,16,power_law_1.2,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,16,power_law_1.2,0.03234559893608093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,16,power_law_1.2,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,16,power_law_1.2,0.03383040130138397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,16,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,16,power_law_1.2,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,16,power_law_1.2,0.041119998693466185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,16,power_law_1.2,0.0490880012512207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,16,power_law_1.2,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,16,power_law_1.2,0.06969599723815918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,16,power_law_1.2,0.09364479780197144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,16,power_law_1.2,0.12001919746398926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,16,power_law_1.2,0.16266239881515504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,16,power_law_1.2,0.19118080139160157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,16,power_law_1.2,0.23326079845428466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,16,power_law_1.2,0.36920320987701416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,16,power_law_1.2,0.7802495956420898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,8,balanced,0.03152533372243246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,8,balanced,0.03387200087308884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,8,balanced,0.03474666674931844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,8,balanced,0.03669866671164831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,8,balanced,0.0424586683511734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,8,balanced,0.0397119993964831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,8,balanced,0.042805333932240806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,8,balanced,0.04233600199222565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,8,balanced,0.04065600037574768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,8,balanced,0.042133331298828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,8,balanced,0.04279999931653341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,8,balanced,0.042634665966033936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,8,balanced,0.044879997769991554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,8,balanced,0.04403733213742574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,8,balanced,0.04631466666857401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,8,balanced,0.04826666911443075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,8,balanced,0.045968001087506614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,8,balanced,0.04906666775544485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,8,balanced,0.050613333781560264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,8,balanced,0.03181333343187968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,8,balanced,0.034458667039871216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,8,balanced,0.05470933516820272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,8,balanced,0.034714666505654655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,8,balanced,0.03676799933115641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,8,balanced,0.059104000528653465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,8,balanced,0.04211199780305227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,8,balanced,0.043247997760772705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,8,balanced,0.042405332128206887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,8,balanced,0.042805333932240806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,8,balanced,0.07113066812356313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,8,balanced,0.04305600126584371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,8,balanced,0.04304533203442892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,8,balanced,0.04285866518815359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,8,balanced,0.043290664752324425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,8,balanced,0.08332799871762593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,8,balanced,0.04528533418973287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,8,balanced,0.044794668753941856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,8,balanced,0.048245335618654885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,8,balanced,0.05034666756788889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,8,balanced,0.10441066821416219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,8,balanced,0.05114666620890299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,8,balanced,0.05283733208974203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,8,balanced,0.05465066432952881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,8,balanced,0.06261866788069408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,8,balanced,0.12800533572832742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,8,balanced,0.06743999818960826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,8,balanced,0.08694932858149211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,8,balanced,0.16446933150291443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,8,balanced,0.1039520005385081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,8,balanced,0.14288000265757242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,8,power_law_1.01,0.03751679956912994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,8,power_law_1.01,0.034668800234794614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,8,power_law_1.01,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,8,balanced,0.21410133441289267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,8,power_law_1.01,0.03642880022525787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,8,power_law_1.01,0.03768959939479828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,8,power_law_1.01,0.04076800048351288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,8,power_law_1.01,0.041657599806785586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,8,balanced,0.17532267173131308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,8,power_law_1.01,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,8,power_law_1.01,0.041459199786186215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,8,power_law_1.01,0.04156799912452698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,8,power_law_1.01,0.04296320080757141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,8,power_law_1.01,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,8,power_law_1.01,0.04359680116176605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,8,power_law_1.01,0.04309119880199432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,8,power_law_1.01,0.04498560130596161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,8,power_law_1.01,0.04683519899845123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,8,balanced,0.25763734181722003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,8,power_law_1.01,0.04819200038909912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,8,balanced,0.23772799968719482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,8,power_law_1.01,0.05187839865684509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,8,power_law_1.01,0.056806397438049314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,8,power_law_1.01,0.06348800063133239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,8,power_law_1.01,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,8,power_law_1.01,0.09026560187339783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,8,power_law_1.01,0.10095360279083251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,8,power_law_1.01,0.13065600395202637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,8,power_law_1.01,0.1668352007865906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,8,power_law_1.01,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,8,power_law_1.01,0.22079999446868898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,8,power_law_1.01,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,8,power_law_1.01,0.32889599800109864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,8,power_law_1.01,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,8,power_law_1.01,0.38933119773864744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,8,power_law_1.01,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,8,balanced,0.3064746658007304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,8,power_law_1.01,0.036774399876594546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,8,power_law_1.01,0.5433343887329102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,8,power_law_1.01,0.040531200170516965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,8,power_law_1.01,1.1119296073913574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,8,power_law_1.01,0.040838399529457094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,8,power_law_1.01,0.04078719913959503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,8,balanced,0.38418134053548175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,8,power_law_1.01,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,8,power_law_1.01,0.04178560078144074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,8,power_law_1.01,0.04248960018157959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,8,power_law_1.01,0.043833601474761966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,8,power_law_1.01,0.04257920086383819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,8,power_law_1.01,0.04537599980831146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,8,power_law_1.01,0.05023999810218811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,8,power_law_1.01,0.04940159916877747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,8,power_law_1.01,0.05098239779472351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,8,power_law_1.01,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,8,power_law_1.01,0.061887997388839724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,8,balanced,0.3744693199793498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,8,power_law_1.01,0.07344639897346497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,8,power_law_1.01,0.0791167974472046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,8,power_law_1.01,0.10853760242462158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,8,power_law_1.01,0.1288831949234009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,8,power_law_1.01,0.17836159467697144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,8,power_law_1.01,0.24158720970153807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,8,power_law_1.01,0.32010879516601565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,8,power_law_1.01,0.417248010635376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,8,power_law_1.01,0.49112958908081056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,8,power_law_1.01,0.753542423248291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,8,power_law_1.01,1.609267234802246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,8,balanced,0.5797333319981893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,8,balanced,0.7160693009694418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,8,balanced,1.1288959980010986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,8,power_law_1.2,0.03591679930686951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,8,power_law_1.2,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,8,power_law_1.2,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,8,power_law_1.2,0.03540480136871338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,8,power_law_1.2,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,8,power_law_1.2,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,8,power_law_1.2,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,8,power_law_1.2,0.041068801283836366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,8,power_law_1.2,0.041459199786186215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,8,power_law_1.2,0.04209280014038086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,8,power_law_1.2,0.04170239865779877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,8,power_law_1.2,0.04261119961738587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,8,power_law_1.2,0.04201599955558777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,8,power_law_1.2,0.043724799156188966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,8,power_law_1.2,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,8,power_law_1.2,0.04707199931144714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,8,power_law_1.2,0.048767998814582825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,8,power_law_1.2,0.05149440169334411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,8,power_law_1.2,0.058911997079849246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,8,power_law_1.2,0.06607360243797303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,8,power_law_1.2,0.07661439776420594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,8,power_law_1.2,0.0984063982963562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,8,power_law_1.2,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,8,power_law_1.2,0.11277439594268798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,8,power_law_1.2,0.03436799943447113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,8,power_law_1.2,0.15387519598007202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,8,power_law_1.2,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,8,power_law_1.2,0.035667198896408084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,8,power_law_1.2,0.18762880563735962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,8,power_law_1.2,0.03731839954853058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,8,power_law_1.2,0.29040639400482177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,8,power_law_1.2,0.04015359878540039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,8,power_law_1.2,0.3894272089004517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,32,balanced,0.04824000100294749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,8,power_law_1.2,0.041196799278259276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,8,power_law_1.2,0.48119678497314455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,32,balanced,0.04671466847260793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,8,power_law_1.2,0.042099198698997496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,32,balanced,0.04738666613896688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,8,power_law_1.2,0.7769472122192382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,8,power_law_1.2,0.04248960018157959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,32,balanced,0.05896000067392985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,8,power_law_1.2,0.041536000370979306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,32,balanced,0.07718400160471599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,8,power_law_1.2,1.5976703643798829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,32,balanced,0.07418666779994965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,8,power_law_1.2,0.04247680008411407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,32,balanced,0.0759200006723404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,8,power_law_1.2,0.04441600143909454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,32,balanced,0.06770133475462596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,8,power_law_1.2,0.044537600874900815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,32,balanced,0.08380800485610962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,8,power_law_1.2,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,32,balanced,0.08682666222254436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,8,power_law_1.2,0.05010560154914856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,32,balanced,0.06957866748174031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,32,balanced,0.07981333136558533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,8,power_law_1.2,0.050988799333572386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,32,balanced,0.07945066690444946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,8,power_law_1.2,0.05185279846191406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,32,balanced,0.07356266677379608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,8,power_law_1.2,0.05617280006408691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,8,power_law_1.2,0.061337602138519284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,32,balanced,0.07918933530648549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,8,power_law_1.2,0.07477759718894958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,32,balanced,0.07196266452471416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,8,power_law_1.2,0.0907904028892517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,32,balanced,0.07523199915885925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,8,power_law_1.2,0.12045439481735229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,32,balanced,0.07356800138950348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,8,power_law_1.2,0.14101760387420653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,8,power_law_1.2,0.18500479459762573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,32,balanced,0.07506133119265239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,8,power_law_1.2,0.23289599418640136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,32,balanced,0.0769760012626648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,8,power_law_1.2,0.4057919979095459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,8,power_law_1.2,0.43578238487243653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,32,balanced,0.08097599943478902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,8,power_law_1.2,0.6246592044830322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,8,power_law_1.2,1.146886444091797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,8,power_law_1.2,2.4673023223876953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,32,balanced,0.10682132840156555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,32,balanced,0.1114453375339508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,32,balanced,0.13275733590126038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,32,balanced,0.15824533502260843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,32,balanced,0.2060319979985555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,32,balanced,0.0401706670721372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,32,balanced,0.04253333310286204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,32,balanced,0.040549332896868386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,32,balanced,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,32,balanced,0.050581331054369606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,32,balanced,0.05855466425418854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,32,balanced,0.05562133093674978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,32,balanced,0.055248002211252846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,32,balanced,0.05542399982611338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,32,balanced,0.056874667604764305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,32,balanced,0.05314133564631144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,32,balanced,0.05459199845790863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,32,balanced,0.053717335065205894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,32,balanced,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,32,balanced,0.05938666562239329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,32,balanced,0.0606826643149058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,32,balanced,0.061018665631612144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,32,balanced,0.06491733094056447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,32,balanced,0.06838933130105336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,32,balanced,0.07682666679223378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,32,balanced,0.08752000331878662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,32,balanced,0.11202667156855266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,32,balanced,0.130703995625178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,32,balanced,0.16523200273513794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,32,balanced,0.04040000090996424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,32,balanced,0.040906667709350586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,32,balanced,0.040922666589419045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,32,balanced,0.04833066463470459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,32,balanced,0.20194133122762045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,32,balanced,0.05904000004132589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,32,balanced,0.05561066667238871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,32,balanced,0.06066666543483734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,32,balanced,0.0569706658522288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,32,balanced,0.05765333275000254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,32,balanced,0.05514133473237356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,32,balanced,0.05895466605822245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,32,balanced,0.05867200096448263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,32,balanced,0.05696000158786774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,32,balanced,0.05500266452630361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,32,balanced,0.27410133679707843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,32,balanced,0.06101333101590475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,32,balanced,0.06327466666698456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,32,balanced,0.06145066519578298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,32,balanced,0.06295999884605408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,32,balanced,0.06651733318964641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,32,balanced,0.0668213317791621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,32,balanced,0.025637333591779072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,32,balanced,0.024453334510326385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,32,balanced,0.02239999920129776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,32,balanced,0.35142401854197186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,32,balanced,0.07318399846553802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,32,balanced,0.02565866708755493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,32,balanced,0.027989332874615986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,32,balanced,0.03234666585922241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,32,balanced,0.04654933512210846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,32,balanced,0.04654400050640106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,32,balanced,0.0935093363126119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,32,balanced,0.0525439977645874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,32,balanced,0.051183998584747314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,32,balanced,0.05045866469542185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,32,balanced,0.04629333317279816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,32,balanced,0.10601600011189778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,32,balanced,0.04324266811211904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,32,balanced,0.03687999894221624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,32,balanced,0.060175999999046326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,32,balanced,0.05797866483529409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,32,balanced,0.05064000189304352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,32,balanced,0.42712001005808514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,32,balanced,0.13144532839457193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,32,balanced,0.05686399837334951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,32,balanced,0.06930666665236156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,32,balanced,0.08551999926567078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,32,balanced,0.157231996456782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,32,balanced,0.10549333691596985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,32,balanced,0.13854933778444925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,32,balanced,0.21046932538350424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,32,balanced,0.1756053368250529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,32,balanced,0.6465280055999756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,32,balanced,0.2452639937400818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,32,balanced,0.26315732796986896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,32,balanced,0.3155946731567383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,32,balanced,0.46371201674143475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,32,balanced,0.3152746756871541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,32,balanced,0.25683732827504474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,32,balanced,0.6095253229141235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,32,balanced,1.2448960145314534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,32,balanced,0.31160000960032147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,32,balanced,0.4692000150680542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,32,balanced,0.7586133480072021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,32,balanced,0.4595946470896403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,32,balanced,1.1991573174794514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,32,balanced,0.886298656463623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,32,balanced,0.861029307047526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,32,balanced,2.3683093388875327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,32,power_law_1.01,0.06320000290870667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,32,power_law_1.01,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,32,power_law_1.01,0.05225600004196167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,32,power_law_1.01,0.058271998167037965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,32,power_law_1.01,0.06816639900207519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,32,power_law_1.01,0.0663424015045166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,32,power_law_1.01,0.0732096016407013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,32,power_law_1.01,0.06890239715576171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,32,power_law_1.01,0.07047039866447449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,32,power_law_1.01,0.07217919826507568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,32,power_law_1.01,0.07228800058364868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,32,power_law_1.01,0.07233279943466187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,32,power_law_1.01,0.07291520237922669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,32,power_law_1.01,0.07340800166130065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,32,power_law_1.01,0.07103360295295716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,32,power_law_1.01,0.0735871970653534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,32,power_law_1.01,0.0851967990398407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,32,power_law_1.01,0.08536959886550903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,32,power_law_1.01,0.10375039577484131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,32,power_law_1.01,0.11772799491882324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,32,power_law_1.01,0.12761600017547609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,32,power_law_1.01,0.17440639734268187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,32,power_law_1.01,0.19906560182571412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,32,power_law_1.01,0.2672703981399536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,32,power_law_1.01,0.35359361171722414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,32,power_law_1.01,0.4553535938262939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,32,power_law_1.01,0.6925568103790283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,32,power_law_1.01,0.7825024127960205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,32,power_law_1.01,1.4431679725646973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,32,power_law_1.01,2.590451240539551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,32,power_law_1.01,0.05441920161247253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,32,power_law_1.01,0.04325119853019714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,32,power_law_1.01,0.04284160137176514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,32,power_law_1.01,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,32,power_law_1.01,0.04741120040416717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,32,power_law_1.01,0.05188480019569397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,32,power_law_1.01,0.05763840079307556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,32,power_law_1.01,0.05432320237159729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,32,power_law_1.01,0.04732159972190857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,32,power_law_1.01,0.05374079942703247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,32,power_law_1.01,0.046387198567390445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,32,power_law_1.01,0.05135999917984009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,32,power_law_1.01,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,32,power_law_1.01,0.05268480181694031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,32,power_law_1.01,0.04915199875831604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,32,power_law_1.01,0.05676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,32,power_law_1.01,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,32,power_law_1.01,0.055820798873901366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,32,power_law_1.01,0.05384960174560547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,32,power_law_1.01,0.055641597509384154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,32,power_law_1.01,0.05548160076141358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,32,power_law_1.01,0.055315202474594115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,32,power_law_1.01,0.053401601314544675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,32,power_law_1.01,0.05749760270118713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,32,power_law_1.01,0.05631359815597534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,32,power_law_1.01,0.06251519918441772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,32,power_law_1.01,0.055270397663116456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,32,power_law_1.01,0.06855040192604064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,32,power_law_1.01,0.055289602279663085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,32,power_law_1.01,0.07989119887351989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,32,power_law_1.01,0.05633280277252197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,32,power_law_1.01,0.08515840172767639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,32,power_law_1.01,0.05728639960289002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,32,power_law_1.01,0.10533759593963624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,32,power_law_1.01,0.059571200609207155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,32,power_law_1.01,0.11580159664154052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,32,power_law_1.01,0.06298239827156067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,32,power_law_1.01,0.15724799633026124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,32,power_law_1.01,0.06535040140151978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,32,power_law_1.01,0.17767679691314697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,32,power_law_1.01,0.07244799733161926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,32,power_law_1.01,0.26502399444580077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.01,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,32,power_law_1.01,0.3309823989868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.01,0.09746559858322143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,32,power_law_1.01,0.4614399909973145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.01,0.10713599920272827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,32,power_law_1.01,0.6059648036956787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.01,0.13451520204544068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,32,power_law_1.01,0.6996992111206055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.01,0.1543552041053772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,32,power_law_1.01,1.220787239074707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.01,0.2189568042755127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,32,power_law_1.01,2.233625602722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.01,0.23726720809936525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.01,0.3877504110336304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.01,0.49679999351501464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,32,power_law_1.01,0.6054399967193603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,32,power_law_1.01,0.9816191673278809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,32,power_law_1.01,1.781260871887207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,32,power_law_1.01,0.0251008003950119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,32,power_law_1.01,0.02242559939622879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,32,power_law_1.01,0.02172800004482269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,32,power_law_1.01,0.02279680073261261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,32,power_law_1.01,0.026464000344276428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,32,power_law_1.01,0.03246079981327057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,32,power_law_1.01,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,32,power_law_1.01,0.044627198576927186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,32,power_law_1.01,0.05237759947776795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,32,power_law_1.01,0.05098239779472351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,32,power_law_1.01,0.05022720098495483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,32,power_law_1.01,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,32,power_law_1.01,0.04783360064029694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,32,power_law_1.01,0.04259839951992035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,32,power_law_1.01,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,32,power_law_1.01,0.04947839975357056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,32,power_law_1.01,0.04560000002384186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,32,power_law_1.01,0.05272960066795349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.01,0.05429760217666626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.01,0.06703360080718994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.01,0.08013439774513245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.01,0.09477120041847228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.01,0.12478079795837402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.01,0.16673920154571534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.01,0.21516799926757812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.01,0.32257280349731443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.01,0.42421760559082033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,32,power_law_1.01,0.47246718406677246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,32,power_law_1.01,0.8447615623474121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,32,power_law_1.01,1.5294719696044923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,32,power_law_1.2,0.06750720143318176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,32,power_law_1.2,0.055289602279663085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,32,power_law_1.2,0.057529598474502563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,32,power_law_1.2,0.06785920262336731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,32,power_law_1.2,0.06955519914627076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,32,power_law_1.2,0.0733568012714386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,32,power_law_1.2,0.07304959893226623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,32,power_law_1.2,0.07528960108757018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,32,power_law_1.2,0.06661760210990905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,32,power_law_1.2,0.07006080150604248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,32,power_law_1.2,0.06946560144424438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,32,power_law_1.2,0.06993280053138733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,32,power_law_1.2,0.07284479737281799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,32,power_law_1.2,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,32,power_law_1.2,0.08012160062789916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,32,power_law_1.2,0.08201599717140198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,32,power_law_1.2,0.08689919710159302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,32,power_law_1.2,0.09772160053253173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,32,power_law_1.2,0.10830719470977783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,32,power_law_1.2,0.1298815965652466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,32,power_law_1.2,0.1535104036331177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,32,power_law_1.2,0.19144320487976074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,32,power_law_1.2,0.2556864023208618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,32,power_law_1.2,0.35136001110076903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,32,power_law_1.2,0.49320321083068847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,32,power_law_1.2,0.7748159885406494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,32,power_law_1.2,0.871827220916748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,32,power_law_1.2,1.2007295608520507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,32,power_law_1.2,1.9839296340942383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,32,power_law_1.2,5.41009292602539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,32,power_law_1.2,0.04833920001983642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,32,power_law_1.2,0.046982398629188536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,32,power_law_1.2,0.042982399463653564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,32,power_law_1.2,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,32,power_law_1.2,0.04438399970531463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,32,power_law_1.2,0.05109120011329651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,32,power_law_1.2,0.05043839812278748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,32,power_law_1.2,0.05411199927330017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,32,power_law_1.2,0.05314559936523437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,32,power_law_1.2,0.05411199927330017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,32,power_law_1.2,0.05459840297698974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,32,power_law_1.2,0.05702400207519531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,32,power_law_1.2,0.05273600220680237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,32,power_law_1.2,0.05477759838104248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,32,power_law_1.2,0.041606399416923526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,32,power_law_1.2,0.05687040090560913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,32,power_law_1.2,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,32,power_law_1.2,0.0597760021686554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,32,power_law_1.2,0.04702720046043396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,32,power_law_1.2,0.061375999450683595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,32,power_law_1.2,0.050361597537994386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,32,power_law_1.2,0.07087360024452209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,32,power_law_1.2,0.05557119846343994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,32,power_law_1.2,0.07891200184822082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,32,power_law_1.2,0.055129599571228025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,32,power_law_1.2,0.05491200089454651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,32,power_law_1.2,0.08762239813804626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,32,power_law_1.2,0.056678402423858645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,32,power_law_1.2,0.11608320474624634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,32,power_law_1.2,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,32,power_law_1.2,0.12722560167312622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,32,power_law_1.2,0.05392000079154968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,32,power_law_1.2,0.15879039764404296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,32,power_law_1.2,0.05589759945869446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,32,power_law_1.2,0.21184000968933106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,32,power_law_1.2,0.056601601839065555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,32,power_law_1.2,0.3125695943832397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,32,power_law_1.2,0.057708799839019775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,32,power_law_1.2,0.40780158042907716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,32,power_law_1.2,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,32,power_law_1.2,0.6438911914825439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,32,power_law_1.2,0.0622655987739563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,32,power_law_1.2,0.7559807777404786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,32,power_law_1.2,0.06686080098152161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,32,power_law_1.2,1.1706879615783692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,32,power_law_1.2,0.07604479789733887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,32,power_law_1.2,1.66046085357666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.2,0.08206719756126404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.2,0.10444799661636353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,32,power_law_1.2,4.485657501220703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.2,0.11676160097122193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.2,0.15529600381851197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.2,0.18662400245666505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.2,0.25708160400390623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.2,0.36801919937133787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.2,0.561510419845581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.2,0.6658624172210693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,32,power_law_1.2,0.023366400599479677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,32,power_law_1.2,0.8937600135803223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,32,power_law_1.2,0.021491199731826782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,32,power_law_1.2,0.021401600539684297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,32,power_law_1.2,1.5575424194335938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,32,power_law_1.2,0.023001599311828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,32,power_law_1.2,3.9414783477783204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,32,power_law_1.2,0.025542399287223815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,32,power_law_1.2,0.033471998572349546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,32,power_law_1.2,0.044607999920845035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,32,power_law_1.2,0.04408319890499115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,32,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,32,power_law_1.2,0.05242239832878113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,32,power_law_1.2,0.049932798743247984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,32,power_law_1.2,0.04922879934310913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,32,power_law_1.2,0.045638400316238406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,32,power_law_1.2,0.042124798893928526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,32,power_law_1.2,0.041222399473190306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,32,power_law_1.2,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,32,power_law_1.2,0.04752640128135681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,32,power_law_1.2,0.05375999808311462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.2,0.05839999914169312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.2,0.06896640062332153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.2,0.07954559922218322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.2,0.10095360279083251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.2,0.127455997467041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.2,0.19769599437713622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.2,0.24019839763641357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.2,0.3376703977584839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.2,0.4508416175842285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,32,power_law_1.2,0.5958335876464844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,32,power_law_1.2,0.9498687744140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,32,power_law_1.2,2.1740928649902345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,16,balanced,0.040405333042144775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,16,balanced,0.04089066634575526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,16,balanced,0.043040002385775246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,16,balanced,0.05495999753475189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,16,balanced,0.06976533432801564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,16,balanced,0.06710400183995564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,16,balanced,0.07738666733105977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,16,balanced,0.06715199848016103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,16,balanced,0.07736533383528392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,16,balanced,0.075162669022878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,16,balanced,0.07065600156784058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,16,balanced,0.07573333382606506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,16,balanced,0.07178666690985362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,16,balanced,0.06919466455777486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,16,balanced,0.07353599866231282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,16,balanced,0.07488533357779185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,16,balanced,0.07293866574764252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,16,balanced,0.07544533411661784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,16,balanced,0.07482133309046428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,16,balanced,0.08161066472530365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,16,balanced,0.08575466275215149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,16,balanced,0.11025599638621013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,16,balanced,0.1199679970741272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,16,balanced,0.14292800426483154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,16,balanced,0.03597866743803024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,16,balanced,0.036703998843828835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,16,balanced,0.040607998768488564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,16,balanced,0.04653333127498627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,16,balanced,0.05277333160241445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,16,balanced,0.05494399865468343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,16,balanced,0.054842665791511536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,16,balanced,0.17330666383107504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,16,balanced,0.05300266544024149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,16,balanced,0.0529013325770696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,16,balanced,0.05443733433882395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,16,balanced,0.03664000084002813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,16,balanced,0.05142400165398916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,16,balanced,0.03658666710058848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,16,balanced,0.061280002196629844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,16,balanced,0.04288533329963684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,16,balanced,0.05482666691144308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,16,balanced,0.050810664892196655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,16,balanced,0.057258665561676025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,16,balanced,0.0613013356924057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,16,balanced,0.060965334375699363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,16,balanced,0.05857066810131073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,16,balanced,0.22780267397562662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,16,balanced,0.05702400207519531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,16,balanced,0.060266668597857155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,16,balanced,0.057061334451039634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,16,balanced,0.06051200131575266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,16,balanced,0.05946666498978933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,16,balanced,0.05672533313433329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,16,balanced,0.06499200065930684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,16,balanced,0.056320001681645714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,16,balanced,0.056346664826075234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,16,balanced,0.06920533378918965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,16,balanced,0.057861333092053734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,16,balanced,0.056832000613212585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,16,balanced,0.08070933322111766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,16,balanced,0.05901333192984263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,16,balanced,0.06113600234190623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,16,balanced,0.09372799595197041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,16,balanced,0.29453333218892414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,16,balanced,0.06316266457239787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,16,balanced,0.06502399841944377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,16,balanced,0.12030399839083354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,16,balanced,0.06765866776307423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,16,balanced,0.07399466633796692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,16,balanced,0.13939199844996134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,16,balanced,0.08045866588751475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,16,balanced,0.10445333520571391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,16,balanced,0.18025066455205283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,16,balanced,0.3498506546020508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,16,balanced,0.1151626706123352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,16,balanced,0.02143999934196472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,16,balanced,0.023056000471115112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,16,balanced,0.022005334496498108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,16,balanced,0.022485333184401195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,16,balanced,0.026122666895389557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,16,balanced,0.14847466349601746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,16,balanced,0.027632000545660656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,16,balanced,0.22222934166590372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,16,balanced,0.03341866781314214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,16,balanced,0.030282666285832722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,16,balanced,0.044810667634010315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,16,balanced,0.04480533301830292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,16,balanced,0.044293334086736046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,16,balanced,0.17786133289337158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,16,balanced,0.050399998823801674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,16,balanced,0.049738665421803795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,16,balanced,0.04517333209514618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,16,balanced,0.0588266650835673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,16,balanced,0.05695466697216034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,16,balanced,0.3075679937998454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,16,balanced,0.05300800005594889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,16,balanced,0.06462400158246358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,16,balanced,0.24417599042256674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,16,balanced,0.5276106595993042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,16,balanced,0.0682239979505539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,16,balanced,0.08501866459846497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,16,balanced,0.0936906635761261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,16,balanced,0.11570666233698527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,16,balanced,0.3945653438568115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,16,balanced,0.3063146670659383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,16,balanced,0.13617066542307535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,16,balanced,0.19448532660802206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,16,balanced,0.24013866980870566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,16,balanced,0.3685226837793986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,16,balanced,0.4818986654281616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,16,balanced,0.3431359926859538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,16,balanced,1.0010666847229004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,16,balanced,0.44489598274230957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,16,balanced,0.5591520071029663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,16,balanced,0.7384533087412516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,16,balanced,0.5500266551971436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,16,balanced,0.8807360331217448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,16,balanced,1.0602453549702961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,16,balanced,1.4274880091349285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,16,balanced,1.732367992401123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,16,power_law_1.01,0.06687999963760376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,16,power_law_1.01,0.05509759783744812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,16,power_law_1.01,0.04696959853172302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,16,power_law_1.01,0.058719998598098753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,16,power_law_1.01,0.061247998476028444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,16,power_law_1.01,0.0697983980178833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,16,power_law_1.01,0.06693120002746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,16,power_law_1.01,0.06935039758682252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,16,power_law_1.01,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,16,power_law_1.01,0.07099519968032837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,16,power_law_1.01,0.06999040246009827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,16,power_law_1.01,0.06865280270576476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,16,power_law_1.01,0.06844800114631652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,16,power_law_1.01,0.06912000179290771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,16,power_law_1.01,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,16,power_law_1.01,0.07416319847106934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,16,power_law_1.01,0.07909119725227357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,16,power_law_1.01,0.08147839903831482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,16,power_law_1.01,0.0972544014453888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,16,power_law_1.01,0.10899200439453124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,16,power_law_1.01,0.12179839611053467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,16,power_law_1.01,0.1484544038772583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,16,power_law_1.01,0.18928639888763427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,16,power_law_1.01,0.2291327953338623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,16,power_law_1.01,0.29605119228363036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,16,power_law_1.01,0.4559232234954834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,16,power_law_1.01,0.5613376140594483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,16,power_law_1.01,0.7225728034973145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,16,power_law_1.01,1.2556927680969239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,16,power_law_1.01,1.7967679977416993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,16,power_law_1.01,0.051622402667999265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,16,power_law_1.01,0.04123519957065582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,16,power_law_1.01,0.04072319865226746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,16,power_law_1.01,0.045126399397850035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,16,power_law_1.01,0.046342399716377256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,16,power_law_1.01,0.05103359818458557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,16,power_law_1.01,0.05178239941596985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,16,power_law_1.01,0.05331839919090271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,16,power_law_1.01,0.05113599896430969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,16,power_law_1.01,0.053414398431777955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,16,power_law_1.01,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,16,power_law_1.01,0.05508480072021484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,16,power_law_1.01,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,16,power_law_1.01,0.05602560043334961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,16,power_law_1.01,0.0591871976852417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,16,power_law_1.01,0.061260801553726194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,16,power_law_1.01,0.06653439998626709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,16,power_law_1.01,0.07768959999084472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,16,power_law_1.01,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,16,power_law_1.01,0.10065280199050904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,16,power_law_1.01,0.11401599645614624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,16,power_law_1.01,0.062015998363494876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,16,power_law_1.01,0.15018240213394166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,16,power_law_1.01,0.05203840136528015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,16,power_law_1.01,0.19532159566879273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,16,power_law_1.01,0.04204800128936768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,16,power_law_1.01,0.243833589553833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,16,power_law_1.01,0.04962559938430786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,16,power_law_1.01,0.0514303982257843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,16,power_law_1.01,0.32695679664611815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,16,power_law_1.01,0.058259201049804685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,16,power_law_1.01,0.4546495914459229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,16,power_law_1.01,0.058355200290679934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,16,power_law_1.01,0.6157567977905274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,16,power_law_1.01,0.05630720257759094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,16,power_law_1.01,0.7492032051086426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,16,power_law_1.01,0.056595200300216676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,16,power_law_1.01,0.021715199947357176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,16,power_law_1.01,1.1854080200195312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,16,power_law_1.01,0.05799040198326111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,16,power_law_1.01,0.0208639994263649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,16,power_law_1.01,0.05687040090560913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,16,power_law_1.01,0.020319999754428865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,16,power_law_1.01,2.2483392715454102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,16,power_law_1.01,0.057715201377868654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,16,power_law_1.01,0.021568000316619873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,16,power_law_1.01,0.05793920159339905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,16,power_law_1.01,0.023231999576091768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,16,power_law_1.01,0.058764797449111936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,16,power_law_1.01,0.02622080147266388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,16,power_law_1.01,0.03269760012626648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,16,power_law_1.01,0.06173440217971802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,16,power_law_1.01,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,16,power_law_1.01,0.06401919722557067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,16,power_law_1.01,0.04360319972038269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,16,power_law_1.01,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,16,power_law_1.01,0.04391680061817169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,16,power_law_1.01,0.0721343994140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,16,power_law_1.01,0.04384639859199524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.01,0.08302720189094544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,16,power_law_1.01,0.0526528000831604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.01,0.09455999732017517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,16,power_law_1.01,0.051481598615646364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.01,0.1059648036956787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,16,power_law_1.01,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.01,0.13134080171585083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,16,power_law_1.01,0.048147198557853696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.01,0.1686527967453003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,16,power_law_1.01,0.048716801404953006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.01,0.2038271903991699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,16,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.01,0.27705600261688235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,16,power_law_1.01,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.01,0.36625919342041013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.01,0.05441280007362366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.01,0.5381247997283936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.01,0.06778879761695862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,16,power_law_1.01,0.605728006362915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.01,0.07311360239982605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,16,power_law_1.01,0.9032192230224609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.01,0.08887680172920227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.01,0.10707199573516846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,16,power_law_1.01,2.0187135696411134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.01,0.14696320295333862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.01,0.17112319469451903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.01,0.2533632040023804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.01,0.32542080879211427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,16,power_law_1.01,0.3760895967483521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,16,power_law_1.01,0.6583551883697509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,16,power_law_1.01,1.2022015571594238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,16,power_law_1.2,0.06926720142364502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,16,power_law_1.2,0.0466623991727829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,16,power_law_1.2,0.04881280064582825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,16,power_law_1.2,0.05366399884223938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,16,power_law_1.2,0.060736000537872314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,16,power_law_1.2,0.07022719979286193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,16,power_law_1.2,0.06702079772949218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,16,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,16,power_law_1.2,0.06785920262336731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,16,power_law_1.2,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,16,power_law_1.2,0.07041280269622803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,16,power_law_1.2,0.0701632022857666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,16,power_law_1.2,0.07248640060424805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,16,power_law_1.2,0.07159039974212647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,16,power_law_1.2,0.07527040243148804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,16,power_law_1.2,0.07671679854393006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,16,power_law_1.2,0.08120319843292237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,16,power_law_1.2,0.09011840224266052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,16,power_law_1.2,0.10021760463714599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,16,power_law_1.2,0.11813119649887086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,16,power_law_1.2,0.13528319597244262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,16,power_law_1.2,0.17242239713668822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,16,power_law_1.2,0.20578560829162598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,16,power_law_1.2,0.3189183950424194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,16,power_law_1.2,0.3831935882568359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,16,power_law_1.2,0.6391295909881591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,16,power_law_1.2,0.6885183811187744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,16,power_law_1.2,0.933619213104248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,16,power_law_1.2,1.866009521484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,16,power_law_1.2,4.432537460327149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,16,power_law_1.2,0.05121279954910278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,16,power_law_1.2,0.03999359905719757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,16,power_law_1.2,0.041280001401901245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,16,power_law_1.2,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,16,power_law_1.2,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,16,power_law_1.2,0.050393599271774295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,16,power_law_1.2,0.05306879878044128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,16,power_law_1.2,0.05367040038108826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,16,power_law_1.2,0.05134720206260681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,16,power_law_1.2,0.05117440223693848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,16,power_law_1.2,0.05141119956970215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,16,power_law_1.2,0.054028797149658206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,16,power_law_1.2,0.05364480018615723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,16,power_law_1.2,0.05731840133666992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,16,power_law_1.2,0.061408001184463504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,16,power_law_1.2,0.06366080045700073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,16,power_law_1.2,0.06876159906387329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,16,power_law_1.2,0.0791808009147644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,16,power_law_1.2,0.06125440001487732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,16,power_law_1.2,0.09419519901275634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,16,power_law_1.2,0.041580799221992495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,16,power_law_1.2,0.10451200008392333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,16,power_law_1.2,0.04268800020217896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,16,power_law_1.2,0.12528640031814575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,16,power_law_1.2,0.04526079893112182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,16,power_law_1.2,0.17998080253601073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,16,power_law_1.2,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,16,power_law_1.2,0.2082304000854492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,16,power_law_1.2,0.05443199872970581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,16,power_law_1.2,0.31475200653076174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,16,power_law_1.2,0.05565440058708191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,16,power_law_1.2,0.02144639939069748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,16,power_law_1.2,0.3993855953216553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,16,power_law_1.2,0.0191551998257637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,16,power_law_1.2,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,16,power_law_1.2,0.5882239818572998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,16,power_law_1.2,0.02025599926710129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,16,power_law_1.2,0.05713919997215271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,16,power_law_1.2,0.8159616470336915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,16,power_law_1.2,0.021241599321365358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,16,power_law_1.2,0.056441599130630495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,16,power_law_1.2,0.9086463928222657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,16,power_law_1.2,0.02173440009355545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,16,power_law_1.2,0.05724160075187683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,16,power_law_1.2,0.058355200290679934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,16,power_law_1.2,0.026444798707962035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,16,power_law_1.2,1.8708223342895507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,16,power_law_1.2,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,16,power_law_1.2,0.0574400007724762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,16,power_law_1.2,3.6640384674072264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,16,power_law_1.2,0.060185599327087405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,16,power_law_1.2,0.03139840066432953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,16,power_law_1.2,0.06244480013847351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,16,power_law_1.2,0.043987199664115906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,16,power_law_1.2,0.064300799369812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,16,power_law_1.2,0.04328320026397705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,16,power_law_1.2,0.07011839747428894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,16,power_law_1.2,0.042822399735450746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,16,power_law_1.2,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,16,power_law_1.2,0.0502016007900238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.2,0.08417279720306396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,16,power_law_1.2,0.052211201190948485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.2,0.10387840270996093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,16,power_law_1.2,0.046988800168037415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.2,0.1239743947982788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,16,power_law_1.2,0.0482367992401123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.2,0.14890240430831908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,16,power_law_1.2,0.04485760033130646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,16,power_law_1.2,0.047251200675964354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.2,0.17875200510025024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.2,0.26720640659332273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,16,power_law_1.2,0.04916479885578155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.2,0.0581055998802185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.2,0.3099071979522705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.2,0.0661184012889862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.2,0.5574592113494873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.2,0.07694720029830933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.2,0.564851188659668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.2,0.09626240134239197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,16,power_law_1.2,0.9475135803222656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.2,0.10914560556411743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,16,power_law_1.2,1.386847972869873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.2,0.1484992027282715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,16,power_law_1.2,2.628153610229492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.2,0.19489279985427857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.2,0.2807168006896973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.2,0.3581696033477783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,16,power_law_1.2,0.42275199890136717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,16,power_law_1.2,0.681766414642334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,16,power_law_1.2,1.5624447822570802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,8,balanced,0.040031999349594116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,8,balanced,0.04132800052563349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,8,balanced,0.043231998880704246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,8,balanced,0.053354665637016296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,8,balanced,0.075573335091273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,8,balanced,0.0697386662165324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,8,balanced,0.06990933418273926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,8,balanced,0.06580266853173573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,8,balanced,0.06999999781449635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,8,balanced,0.0712960014740626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,8,balanced,0.06774400174617767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,8,balanced,0.06786666810512543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,8,balanced,0.06670400003592174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,8,balanced,0.06876799960931142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,8,balanced,0.07441600163777669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,8,balanced,0.07250133156776428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,8,balanced,0.0758240024248759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,8,balanced,0.07833600044250488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,8,balanced,0.08262933293978374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,8,balanced,0.08919999996821086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,8,balanced,0.09434666236241658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,8,balanced,0.03572800010442734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,8,balanced,0.03756266583998998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,8,balanced,0.03898133337497711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,8,balanced,0.04288533329963684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,8,balanced,0.12406933307647705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,8,balanced,0.054272000988324486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,8,balanced,0.05533866584300995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,8,balanced,0.05319466690222422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,8,balanced,0.053957333167394005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,8,balanced,0.1348426640033722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,8,balanced,0.054117331902186074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,8,balanced,0.05643199880917867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,8,balanced,0.05331199864546458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,8,balanced,0.054117331902186074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,8,balanced,0.056976000467936196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,8,balanced,0.05571199953556061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,8,balanced,0.1694773236910502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,8,balanced,0.061162665486335754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,8,balanced,0.06318933268388112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,8,balanced,0.06494933366775513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,8,balanced,0.07176533341407776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,8,balanced,0.2097813288370768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,8,balanced,0.07758399844169617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,8,balanced,0.09372267127037048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,8,balanced,0.10348266363143921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,8,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,8,balanced,0.040378667414188385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,8,balanced,0.04474666714668274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,8,balanced,0.051498666405677795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,8,balanced,0.138672004143397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,8,balanced,0.0653653343518575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,8,balanced,0.28383467594782513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,8,balanced,0.06401599943637848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,8,balanced,0.0647680014371872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,8,balanced,0.01951466624935468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,8,balanced,0.06571199993292491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,8,balanced,0.02203733225663503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,8,balanced,0.15843733151753744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,8,balanced,0.06557333469390869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,8,balanced,0.02213866760333379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,8,balanced,0.06369066735108693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,8,balanced,0.02236266682545344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,8,balanced,0.02701866626739502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,8,balanced,0.06106133262316386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,8,balanced,0.02622399975856145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,8,balanced,0.06306133170922597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,8,balanced,0.030400000512599945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,8,balanced,0.06486399968465169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,8,balanced,0.2120373249053955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,8,balanced,0.029109333952267964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,8,balanced,0.06297066807746887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,8,balanced,0.03856533269087473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,8,balanced,0.06896000107129414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,8,balanced,0.0366293340921402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,8,balanced,0.36689599355061847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,8,balanced,0.03446399917205175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,8,balanced,0.0689279983441035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,8,balanced,0.0484746644894282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,8,balanced,0.07102933526039124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,8,balanced,0.26872533559799194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,8,balanced,0.05007466673851013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,8,balanced,0.07542933523654938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,8,balanced,0.04845866560935974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,8,balanced,0.05890133480230967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,8,balanced,0.07863999903202057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,8,balanced,0.059215997656186424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,8,balanced,0.0881226658821106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,8,balanced,0.05579199890295664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,8,balanced,0.06541866560777028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,8,balanced,0.0986400047938029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,8,balanced,0.06891199946403503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,8,balanced,0.37559465567270917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,8,balanced,0.13024000326792398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,8,balanced,0.08830400307973225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,8,balanced,0.4439573287963867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,8,balanced,0.09032533566157024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,8,balanced,0.14832533399264017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,8,balanced,0.11053867141405742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,8,balanced,0.4867839813232422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,8,balanced,0.1920586625734965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,8,balanced,0.12734933694203696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,8,balanced,0.23675199349721274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,8,balanced,0.18045334021250406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,8,balanced,0.5963306824366251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,8,balanced,0.6764000256856283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,8,balanced,0.21267733971277872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,8,balanced,0.3298986752827962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,8,balanced,0.3009386658668518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,8,balanced,0.4246559937795003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,8,balanced,0.9214239915211996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,8,balanced,0.3885386784871419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,8,balanced,0.5175360043843588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,8,balanced,0.4757386843363444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,8,balanced,1.2827839851379395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,8,balanced,0.7998773256937662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,8,balanced,1.780373255411784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,8,balanced,0.7612373034159342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,8,balanced,1.529904047648112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,8,balanced,1.4902186393737793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,8,power_law_1.01,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,8,power_law_1.01,0.05012480020523071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,8,power_law_1.01,0.048467200994491574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,8,power_law_1.01,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,8,power_law_1.01,0.05711359977722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,8,power_law_1.01,0.06425600051879883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,8,power_law_1.01,0.06698240041732788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,8,power_law_1.01,0.06595839858055115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,8,power_law_1.01,0.06877440214157104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,8,power_law_1.01,0.04425599873065948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,8,power_law_1.01,0.06993280053138733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,8,power_law_1.01,0.03868159949779511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,8,power_law_1.01,0.06752640008926392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,8,power_law_1.01,0.03943040072917938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,8,power_law_1.01,0.06986879706382751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,8,power_law_1.01,0.043968001008033754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,8,power_law_1.01,0.07033600211143494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,8,power_law_1.01,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,8,power_law_1.01,0.052934402227401735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,8,power_law_1.01,0.07120640277862549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,8,power_law_1.01,0.051020801067352295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,8,power_law_1.01,0.07430400252342224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,8,power_law_1.01,0.05310720205307007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,8,power_law_1.01,0.07587839961051941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,8,power_law_1.01,0.05369600057601929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,8,power_law_1.01,0.0761023998260498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,8,power_law_1.01,0.055129599571228025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,8,power_law_1.01,0.08519039750099182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,8,power_law_1.01,0.05548160076141358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,8,power_law_1.01,0.09215360283851623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,8,power_law_1.01,0.054016000032424925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,8,power_law_1.01,0.11110399961471558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,8,power_law_1.01,0.05697280168533325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,8,power_law_1.01,0.12316160202026367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,8,power_law_1.01,0.05886080265045166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,8,power_law_1.01,0.1480191946029663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,8,power_law_1.01,0.06345599889755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,8,power_law_1.01,0.18380160331726075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,8,power_law_1.01,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,8,power_law_1.01,0.23758080005645751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,8,power_law_1.01,0.07066239714622498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,8,power_law_1.01,0.3064896106719971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,8,power_law_1.01,0.07958400249481201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,8,power_law_1.01,0.41393280029296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,8,power_law_1.01,0.09244800209999085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,8,power_law_1.01,0.580460786819458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,8,power_law_1.01,0.10676480531692505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,8,power_law_1.01,0.6961088180541992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,8,power_law_1.01,0.12969599962234496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,8,power_law_1.01,1.1526528358459474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,8,power_law_1.01,0.1595199942588806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,8,power_law_1.01,0.19816960096359254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,8,power_law_1.01,2.129318428039551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,8,power_law_1.01,0.281823992729187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,8,power_law_1.01,0.33909759521484373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,8,power_law_1.01,0.5266880035400391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,8,power_law_1.01,0.7395967960357666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,8,power_law_1.01,0.7557184219360351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,8,power_law_1.01,1.3203136444091796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,8,power_law_1.01,2.6193471908569337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,8,power_law_1.01,0.04983679950237274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,8,power_law_1.01,0.0420415997505188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,8,power_law_1.01,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,8,power_law_1.01,0.05125120282173157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,8,power_law_1.01,0.054079997539520266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,8,power_law_1.01,0.0651968002319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,8,power_law_1.01,0.06164479851722717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,8,power_law_1.01,0.06220800280570984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,8,power_law_1.01,0.06230400204658508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,8,power_law_1.01,0.06443520188331604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,8,power_law_1.01,0.06260480284690857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,8,power_law_1.01,0.0651968002319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,8,power_law_1.01,0.06528000235557556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,8,power_law_1.01,0.06513280272483826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,8,power_law_1.01,0.07000960111618042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,8,power_law_1.01,0.01883520036935806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,8,power_law_1.01,0.0707584023475647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,8,power_law_1.01,0.01990399956703186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,8,power_law_1.01,0.07489920258522034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,8,power_law_1.01,0.020019200444221497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,8,power_law_1.01,0.01966080069541931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,8,power_law_1.01,0.0862272024154663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,8,power_law_1.01,0.021875199675559998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.01,0.09371520280838012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,8,power_law_1.01,0.024864000082015992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.01,0.10705920457839965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,8,power_law_1.01,0.027551999688148497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.01,0.12197760343551636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,8,power_law_1.01,0.02619520127773285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.01,0.1564288020133972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,8,power_law_1.01,0.03208959996700287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.01,0.20188798904418945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,8,power_law_1.01,0.031974399089813234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.01,0.26433279514312746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,8,power_law_1.01,0.03182719945907593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.01,0.32520959377288816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,8,power_law_1.01,0.04615679979324341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.01,0.4859327793121338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,8,power_law_1.01,0.04642559885978699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.01,0.6272448062896728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,8,power_law_1.01,0.04760960042476654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,8,power_law_1.01,0.7750016212463379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,8,power_law_1.01,0.052697598934173584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,8,power_law_1.01,1.1854592323303224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,8,power_law_1.01,0.052832001447677614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,8,power_law_1.01,0.0521664023399353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,8,power_law_1.01,2.243724822998047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,8,power_law_1.01,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.01,0.053491199016571046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.01,0.06570240259170532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.01,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.01,0.08970879912376403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.01,0.10423680543899536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.01,0.1373695969581604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.01,0.17136000394821166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.01,0.23876481056213378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.01,0.3027776002883911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,8,power_law_1.01,0.380128002166748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,8,power_law_1.01,0.5544064044952393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,8,power_law_1.01,1.1607487678527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,8,power_law_1.2,0.053311997652053834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,8,power_law_1.2,0.045817598700523376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,8,power_law_1.2,0.048614400625228885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,8,power_law_1.2,0.05253120064735413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,8,power_law_1.2,0.053913599252700804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,8,power_law_1.2,0.04533120095729828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,8,power_law_1.2,0.06644480228424073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,8,power_law_1.2,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,8,power_law_1.2,0.06783999800682068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,8,power_law_1.2,0.04095360040664673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,8,power_law_1.2,0.06960639953613282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,8,power_law_1.2,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,8,power_law_1.2,0.06886399984359741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,8,power_law_1.2,0.04460160136222839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,8,power_law_1.2,0.06934400200843811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,8,power_law_1.2,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,8,power_law_1.2,0.07061120271682739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,8,power_law_1.2,0.05231999754905701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,8,power_law_1.2,0.06908800005912781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,8,power_law_1.2,0.05249919891357422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,8,power_law_1.2,0.06874240040779114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,8,power_law_1.2,0.05375360250473023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,8,power_law_1.2,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,8,power_law_1.2,0.054745602607727054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,8,power_law_1.2,0.07785599827766418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,8,power_law_1.2,0.0773311972618103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,8,power_law_1.2,0.05589119791984558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,8,power_law_1.2,0.08151040077209473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,8,power_law_1.2,0.05567359924316406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,8,power_law_1.2,0.08848000168800355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,8,power_law_1.2,0.056908798217773435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,8,power_law_1.2,0.0934656023979187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,8,power_law_1.2,0.05914880037307739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,8,power_law_1.2,0.11631360054016113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,8,power_law_1.2,0.06478719711303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,8,power_law_1.2,0.1361024022102356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,8,power_law_1.2,0.06980479955673217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,8,power_law_1.2,0.07388160228729249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,8,power_law_1.2,0.16392960548400878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,8,power_law_1.2,0.0828927993774414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,8,power_law_1.2,0.19356160163879393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,8,power_law_1.2,0.0945792019367218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,8,power_law_1.2,0.29288320541381835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,8,power_law_1.2,0.11412479877471923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,8,power_law_1.2,0.3961535930633545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,8,power_law_1.2,0.1322816014289856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,8,power_law_1.2,0.4720191955566406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,8,power_law_1.2,0.1702463984489441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,8,power_law_1.2,0.7092160224914551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,8,power_law_1.2,0.21369600296020508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,8,power_law_1.2,0.8359999656677246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,8,power_law_1.2,0.306060791015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,8,power_law_1.2,1.4779199600219726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,8,power_law_1.2,0.38869760036468504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,8,power_law_1.2,2.7294591903686523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,8,power_law_1.2,0.5852223873138428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,8,power_law_1.2,0.7846720218658447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,8,power_law_1.2,1.006169605255127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,8,power_law_1.2,1.6083648681640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,8,power_law_1.2,3.254848098754883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,8,power_law_1.2,0.05271040201187134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,8,power_law_1.2,0.04184960126876831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,8,power_law_1.2,0.04382080137729645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,8,power_law_1.2,0.04596480131149292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,8,power_law_1.2,0.04924800097942352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,8,power_law_1.2,0.06351360082626342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,8,power_law_1.2,0.06324480175971985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,8,power_law_1.2,0.06247680187225342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,8,power_law_1.2,0.0642304003238678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,8,power_law_1.2,0.06269440054893494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,8,power_law_1.2,0.0648576021194458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,8,power_law_1.2,0.06493440270423889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,8,power_law_1.2,0.01937279999256134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,8,power_law_1.2,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,8,power_law_1.2,0.01969279944896698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,8,power_law_1.2,0.06665599942207337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,8,power_law_1.2,0.019577600061893463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,8,power_law_1.2,0.07335039973258972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,8,power_law_1.2,0.020051200687885285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,8,power_law_1.2,0.0743552029132843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,8,power_law_1.2,0.02133760005235672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,8,power_law_1.2,0.0796992003917694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,8,power_law_1.2,0.02655999958515167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,8,power_law_1.2,0.08950399756431579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,8,power_law_1.2,0.026636800169944762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.2,0.09757440090179444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,8,power_law_1.2,0.026815998554229736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.2,0.11858559846878051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,8,power_law_1.2,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.2,0.13224960565567018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,8,power_law_1.2,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.2,0.16780799627304077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,8,power_law_1.2,0.031999999284744264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.2,0.2111423969268799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,8,power_law_1.2,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.2,0.2765887975692749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,8,power_law_1.2,0.04656639993190766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.2,0.38618879318237304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,8,power_law_1.2,0.046323201060295104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.2,0.5530367851257324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,8,power_law_1.2,0.05132799744606018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.2,0.7420800209045411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,8,power_law_1.2,0.05284479856491089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,8,power_law_1.2,0.8789248466491699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,8,power_law_1.2,0.048691201210021975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,8,power_law_1.2,1.454963207244873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,8,power_law_1.2,0.04899199903011322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.2,0.05738880038261414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,8,power_law_1.2,3.112607955932617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.2,0.06802560091018676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.2,0.07522559762001038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.2,0.09107199907302857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.2,0.11082240343093872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.2,0.1507840037345886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.2,0.18874239921569824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.2,0.24447360038757324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.2,0.3396672010421753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,8,power_law_1.2,0.3844095945358276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,8,power_law_1.2,0.6656576156616211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,8,power_law_1.2,1.297612762451172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,4,balanced,0.039813332259655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,4,balanced,0.04298133154710134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,4,balanced,0.04548799991607666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,4,balanced,0.05552533268928528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,4,balanced,0.07447466750939687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,4,balanced,0.03606933355331421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,4,balanced,0.07563733557860057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,4,balanced,0.03881066789229711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,4,balanced,0.07297066847483318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,4,balanced,0.04264000058174133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,4,balanced,0.07220800220966339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,4,balanced,0.04824000100294749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,4,balanced,0.0709386666615804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,4,balanced,0.06343466540177663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,4,balanced,0.07187733550866444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,4,balanced,0.06452799836794536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,4,balanced,0.07460266848405202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,4,balanced,0.062165334820747375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,4,balanced,0.07366933425267537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,4,balanced,0.0621013343334198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,4,balanced,0.07428800066312154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,4,balanced,0.06313066681226094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,4,balanced,0.07392533123493195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,4,balanced,0.06155199805895487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,4,balanced,0.060549333691596985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,4,balanced,0.07909866670767467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,4,balanced,0.06275199850400288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,4,balanced,0.08137066662311554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,4,balanced,0.06528000036875407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,4,balanced,0.08123200138409932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,4,balanced,0.06353599826494853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,4,balanced,0.0690773328145345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,4,balanced,0.08904000123341878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,4,balanced,0.0713973343372345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,4,balanced,0.09252267082532246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,4,balanced,0.07724266747633617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,4,balanced,0.08897599577903748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,4,balanced,0.10525866349538167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,4,balanced,0.09529067079226176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,4,balanced,0.11423466602961223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,4,balanced,0.11426132917404175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,4,balanced,0.15429866313934326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,4,balanced,0.12780800461769104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,4,balanced,0.17594667275746664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,4,balanced,0.17520000537236533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,4,balanced,0.20363199710845947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,4,balanced,0.23177599906921387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,4,balanced,0.2825813293457031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,4,balanced,0.2922186652819316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,4,balanced,0.36369065443674725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,4,balanced,0.40425066153208417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,4,balanced,0.5237813393274943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,4,power_law_1.01,0.04625920057296753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,4,balanced,0.5212159951527914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,4,power_law_1.01,0.04519680142402649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,4,power_law_1.01,0.0458624005317688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,4,power_law_1.01,0.04819200038909912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,4,power_law_1.01,0.03860479891300202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,4,balanced,0.6818506717681885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,4,power_law_1.01,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,4,power_law_1.01,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,4,power_law_1.01,0.05676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,4,power_law_1.01,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,4,power_law_1.01,0.06944000124931335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,4,power_law_1.01,0.05119360089302063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,4,power_law_1.01,0.0709119975566864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,4,power_law_1.01,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,4,power_law_1.01,0.06024320125579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,4,power_law_1.01,0.07288960218429566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,4,power_law_1.01,0.06061440110206604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,4,power_law_1.01,0.07043200135231018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,4,power_law_1.01,0.06284160017967225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,4,power_law_1.01,0.07125120162963867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,4,power_law_1.01,0.0648256003856659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,4,power_law_1.01,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,4,power_law_1.01,0.06275839805603027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,4,power_law_1.01,0.0728384017944336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,4,power_law_1.01,0.06641280055046081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,4,power_law_1.01,0.07559040188789368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,4,balanced,0.6290506521860758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,4,power_law_1.01,0.06793599724769592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,4,power_law_1.01,0.07530239820480347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,4,power_law_1.01,0.06899840235710145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,4,balanced,0.8325440088907877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,4,power_law_1.01,0.08127999901771546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,4,power_law_1.01,0.0737280011177063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,4,power_law_1.01,0.08048639893531799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,4,power_law_1.01,0.0775168001651764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,4,power_law_1.01,0.08492159843444824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,4,power_law_1.01,0.08121600151062011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,4,power_law_1.01,0.09383040070533752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,4,power_law_1.01,0.09726719856262207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,4,power_law_1.01,0.10487040281295776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,4,power_law_1.01,0.10429439544677735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,4,power_law_1.01,0.12488960027694702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,4,power_law_1.01,0.132915198802948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,4,power_law_1.01,0.14174720048904418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,4,power_law_1.01,0.15993599891662597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,4,power_law_1.01,0.18258559703826904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,4,power_law_1.01,0.19338879585266114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,4,power_law_1.01,0.2279360055923462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,4,power_law_1.01,0.24068479537963866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,4,power_law_1.01,0.31088640689849856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,4,power_law_1.01,0.33231360912323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,4,power_law_1.01,0.38134400844573973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,4,power_law_1.01,0.459065580368042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,4,power_law_1.01,0.5106112003326416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,4,power_law_1.01,0.6330687999725342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,4,power_law_1.01,0.6860799789428711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,4,power_law_1.01,0.7969088077545166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,4,power_law_1.01,0.7888832092285156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,4,power_law_1.01,0.9917759895324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,4,power_law_1.01,1.2136320114135741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,4,power_law_1.01,1.564121627807617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,4,power_law_1.01,2.4547584533691404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,4,balanced,1.2965599695841472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,4,balanced,0.9622933069864908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,4,power_law_1.01,3.212895965576172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,4,balanced,2.5303093592325845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,4,balanced,1.8607199986775715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,4,power_law_1.2,0.04302720129489899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,4,power_law_1.2,0.039263999462127684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,4,power_law_1.2,0.04069760143756866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,4,power_law_1.2,0.04540160000324249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,4,power_law_1.2,0.050246399641036985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,4,power_law_1.2,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,4,power_law_1.2,0.06040319800376892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,4,power_law_1.2,0.06149759888648987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,4,power_law_1.2,0.046035200357437134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,4,power_law_1.2,0.06306560039520263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,4,power_law_1.2,0.04746240079402923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,4,power_law_1.2,0.06095359921455383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,4,power_law_1.2,0.05429760217666626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,4,power_law_1.2,0.06251519918441772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,4,power_law_1.2,0.05678079724311828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,4,power_law_1.2,0.061440002918243405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,4,power_law_1.2,0.06812160015106201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,4,power_law_1.2,0.06636800169944763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,4,power_law_1.2,0.06952959895133973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,4,power_law_1.2,0.06614400148391723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,4,power_law_1.2,0.07155200242996215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,4,power_law_1.2,0.06758400201797485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,4,power_law_1.2,0.07171199917793274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,4,power_law_1.2,0.07461760044097901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,4,power_law_1.2,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,4,power_law_1.2,0.07842559814453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,4,power_law_1.2,0.0714303970336914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,4,power_law_1.2,0.08529919981956482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,4,power_law_1.2,0.07266560196876526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,4,power_law_1.2,0.09923200011253357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,4,power_law_1.2,0.07389439940452576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,4,power_law_1.2,0.10869760513305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,4,power_law_1.2,0.0772159993648529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,4,power_law_1.2,0.12849279642105102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,4,power_law_1.2,0.0803391993045807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,4,power_law_1.2,0.15259519815444947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,4,power_law_1.2,0.08349440097808838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,4,power_law_1.2,0.2169856071472168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,4,power_law_1.2,0.08532480001449586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,4,power_law_1.2,0.25731840133666994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,4,power_law_1.2,0.09868800044059753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,4,power_law_1.2,0.3419264078140259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,4,power_law_1.2,0.1043328046798706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,4,power_law_1.2,0.4733119964599609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,4,power_law_1.2,0.12627840042114258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,4,power_law_1.2,0.14911359548568726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,4,power_law_1.2,0.7413695812225342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,4,power_law_1.2,0.18084479570388795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,4,power_law_1.2,1.012883186340332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,4,power_law_1.2,0.22884480953216552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,4,power_law_1.2,1.265561580657959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,4,power_law_1.2,0.2952831983566284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,4,power_law_1.2,1.9408447265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,4,power_law_1.2,0.4017216205596924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,4,power_law_1.2,4.301907348632812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,4,power_law_1.2,0.5935935974121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,2,balanced,0.04019733270009359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,4,power_law_1.2,0.7468031883239746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,2,balanced,0.0435146689414978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,2,balanced,0.04667733112970988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,4,power_law_1.2,1.0080191612243652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,2,balanced,0.05778666834036509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,4,power_law_1.2,1.4769984245300294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,2,balanced,0.07576533158620198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,2,balanced,0.04253333310286204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,2,balanced,0.07815466821193695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,2,balanced,0.04764799773693085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,4,power_law_1.2,3.2790847778320313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,2,balanced,0.0753600001335144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,2,balanced,0.052933335304260254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,2,balanced,0.0759200006723404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,2,balanced,0.06597866614659627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,2,balanced,0.08032000064849854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,2,balanced,0.09160000085830688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,2,balanced,0.07545066873232524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,2,balanced,0.08839999636014302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,2,balanced,0.07871466875076294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,2,balanced,0.08970666925112407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,2,balanced,0.08138133088747661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,2,balanced,0.08846933643023173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,2,balanced,0.08688533306121826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,2,balanced,0.08154666423797607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,2,balanced,0.08779733379681905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,2,balanced,0.08351999521255493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,2,balanced,0.08761599659919739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,2,balanced,0.09261866410573323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,2,balanced,0.09269332885742188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,2,balanced,0.09457600116729736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,2,balanced,0.09391466776529948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,2,balanced,0.10034666458765666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,2,balanced,0.09175999959309895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,2,balanced,0.11688533425331116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,2,balanced,0.09758399923642476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,2,balanced,0.10038933157920837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,2,balanced,0.12593600153923035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,2,balanced,0.10362133383750916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,2,balanced,0.15131200353304544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,2,balanced,0.11332799990971883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,2,balanced,0.17229332526524863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,2,balanced,0.12045866250991821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,2,balanced,0.1377066671848297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,2,balanced,0.24977066119511923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,2,balanced,0.15664533774058023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,2,balanced,0.2967306574185689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,2,balanced,0.22410666942596436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,2,balanced,0.4283999999364217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,2,balanced,0.2653706669807434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,2,balanced,0.5584106842676798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,2,balanced,0.3708159923553467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,2,balanced,0.4768799940745036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,2,balanced,0.808784008026123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,2,balanced,0.6786826451619467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,2,balanced,1.055615981419881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,2,balanced,0.8724799950917562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,2,balanced,1.3075786431630452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,2,balanced,1.0794080098470051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,2,balanced,2.054687976837158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,2,power_law_1.01,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,2,power_law_1.01,0.04447360038757324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,2,power_law_1.01,0.049542400240898135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,2,power_law_1.01,0.057708799839019775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,2,power_law_1.01,0.06458879709243774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,2,power_law_1.01,0.0831488013267517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,2,power_law_1.01,0.04283519983291626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,2,power_law_1.01,0.08266239762306213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,2,power_law_1.01,0.04258559942245484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,2,power_law_1.01,0.08652799725532531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,2,power_law_1.01,0.08698239922523499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,2,power_law_1.01,0.04357759952545166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,2,power_law_1.01,0.08607360124588012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,2,power_law_1.01,0.051641601324081424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,2,balanced,1.6999732653299968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,2,power_law_1.01,0.08786560297012329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,2,power_law_1.01,0.058284801244735715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,2,power_law_1.01,0.09200639724731445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,2,power_law_1.01,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,2,power_law_1.01,0.09123839735984803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,2,power_law_1.01,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,2,power_law_1.01,0.09269760251045227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,2,power_law_1.01,0.07395200133323669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,2,power_law_1.01,0.09784319996833801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,2,power_law_1.01,0.073580801486969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,2,power_law_1.01,0.09971200227737427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,2,power_law_1.01,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,2,power_law_1.01,0.10682879686355591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,2,power_law_1.01,0.07811200022697448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,2,power_law_1.01,0.12139519453048705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,2,power_law_1.01,0.08120959997177124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,2,power_law_1.01,0.08369280099868774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,2,power_law_1.01,0.13935359716415405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,2,power_law_1.01,0.08401280045509338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,2,power_law_1.01,0.16261759996414185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,2,power_law_1.01,0.09105920195579528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,2,power_law_1.01,0.1837056040763855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,2,power_law_1.01,0.09578880071640014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,2,power_law_1.01,0.23695359230041504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,2,power_law_1.01,0.1050879955291748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,2,power_law_1.01,0.30670719146728515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,2,power_law_1.01,0.11738239526748658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,2,power_law_1.01,0.41580801010131835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,2,power_law_1.01,0.14180480241775512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,2,power_law_1.01,0.5370944023132325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,2,power_law_1.01,0.1709247946739197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,2,power_law_1.01,0.7892735958099365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,2,power_law_1.01,0.20337920188903807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,2,power_law_1.01,1.0041024208068847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,2,power_law_1.01,0.26074240207672117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,2,power_law_1.01,1.2737088203430176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,2,power_law_1.01,0.3287103891372681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,2,power_law_1.01,1.9373760223388672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,2,power_law_1.01,0.45357441902160645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,2,balanced,4.058650652567546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,2,power_law_1.01,0.5925888061523438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,2,power_law_1.01,4.004595184326172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,2,power_law_1.01,0.8799551963806153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,2,power_law_1.01,1.1719552040100099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,2,power_law_1.01,1.526841640472412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,2,power_law_1.01,2.3516351699829103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,2,power_law_1.01,4.682751846313477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,2,balanced,3.3457813262939453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,2,power_law_1.2,0.045388799905776975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,2,power_law_1.2,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,2,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,2,power_law_1.2,0.043007999658584595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,2,power_law_1.2,0.05812479853630066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,2,power_law_1.2,0.041177600622177124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,2,power_law_1.2,0.06447359919548035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,2,power_law_1.2,0.044486400485038755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,2,power_law_1.2,0.0837823987007141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,2,power_law_1.2,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,2,power_law_1.2,0.08269439935684204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,2,power_law_1.2,0.05504000186920166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,2,power_law_1.2,0.085971200466156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,2,power_law_1.2,0.07021440267562866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,2,power_law_1.2,0.08641279935836792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,2,power_law_1.2,0.07311360239982605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,2,power_law_1.2,0.0738431990146637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,2,power_law_1.2,0.08531200289726257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,1,balanced,0.05031999945640564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,1,balanced,0.052576000491778054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,2,power_law_1.2,0.07685760259628296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,2,power_law_1.2,0.08605440258979798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,1,balanced,0.06004266440868378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,2,power_law_1.2,0.09069439768791199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,2,power_law_1.2,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,1,balanced,0.07680533329645793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,2,power_law_1.2,0.09220479726791382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,2,power_law_1.2,0.07956479787826538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,1,balanced,0.10800533493359883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,2,power_law_1.2,0.09173759818077087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,2,power_law_1.2,0.08133119940757752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,1,balanced,0.11151466766993205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,2,power_law_1.2,0.09811840057373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,2,power_law_1.2,0.0819711983203888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,1,balanced,0.11089600125948589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,1,balanced,0.048341333866119385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,2,power_law_1.2,0.10209920406341552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,2,power_law_1.2,0.08773760199546814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,1,balanced,0.11191999912261963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,1,balanced,0.05157866577307383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,2,power_law_1.2,0.10776319503784179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,1,balanced,0.11171733339627583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,2,power_law_1.2,0.09552000164985656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,1,balanced,0.05958933134873708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,1,balanced,0.11390933394432068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,1,balanced,0.07691200077533722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,2,power_law_1.2,0.12278399467468262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,2,power_law_1.2,0.09741439819335937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,1,balanced,0.10294933120409648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,1,balanced,0.11276266972223918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,2,power_law_1.2,0.1427199959754944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,2,power_law_1.2,0.10743039846420288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,1,balanced,0.10322133700052898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,1,balanced,0.11643200119336446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,2,power_law_1.2,0.17853440046310426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,2,power_law_1.2,0.12553600072860718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,1,balanced,0.1065120001633962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,1,balanced,0.1165066659450531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,2,power_law_1.2,0.20177919864654542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,2,power_law_1.2,0.13834240436553955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,1,balanced,0.10748799641927083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,1,balanced,0.1218773325284322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,2,power_law_1.2,0.26408960819244387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,1,balanced,0.10718933741251628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,2,power_law_1.2,0.17849600315093994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,1,balanced,0.12770666678746542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,1,balanced,0.10691199700037639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,2,power_law_1.2,0.3160128116607666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,2,power_law_1.2,0.19928319454193116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,1,balanced,0.11115733782450359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,1,balanced,0.12895466883977255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,2,power_law_1.2,0.44312319755554197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,2,power_law_1.2,0.2781375885009766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,1,balanced,0.1143999993801117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,2,power_law_1.2,0.3337023973464966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,2,power_law_1.2,0.5678976058959961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,1,balanced,0.1374506652355194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,1,balanced,0.11551466584205627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,2,power_law_1.2,0.5194752216339111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,2,power_law_1.2,0.8237440109252929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,1,balanced,0.1549493372440338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,1,balanced,0.11993066469828288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,2,power_law_1.2,0.620684814453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,2,power_law_1.2,1.2478528022766113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,1,balanced,0.1288053294022878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,2,power_law_1.2,0.9339263916015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,1,balanced,0.1721280018488566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,2,power_law_1.2,1.3445119857788086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,1,balanced,0.13276799519856772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,2,power_law_1.2,1.2550527572631835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,2,power_law_1.2,2.176665687561035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,1,balanced,0.142767995595932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,1,balanced,0.20919466018676758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,2,power_law_1.2,1.4884991645812988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,2,power_law_1.2,4.307648086547852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,1,balanced,0.16548267006874084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,2,power_law_1.2,2.5937280654907227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,1,balanced,0.24273600180943808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,1,balanced,0.1846933364868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,2,power_law_1.2,4.790483093261718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,1,balanced,0.36655465761820477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,1,balanced,0.22744532426198324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,1,balanced,0.2704266707102458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,1,balanced,0.43822399775187176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,1,balanced,0.40465601285298664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,1,balanced,0.6217653354008993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,1,balanced,0.497109333674113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,1,balanced,0.7969280083974203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,1,balanced,0.7291946411132812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,1,balanced,0.9504373073577881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,1,balanced,1.1676692962646484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,1,balanced,1.4072480201721191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,1,balanced,1.5060853958129883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,1,balanced,1.8288586934407551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,1,balanced,1.8742879231770833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,1,balanced,2.2782719930013022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,1,balanced,2.9176534016927085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,1,balanced,3.586597442626953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,1,power_law_1.01,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,1,power_law_1.01,0.04826239943504333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,1,power_law_1.01,0.05634559988975525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,1,power_law_1.01,0.06755200028419495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,1,power_law_1.01,0.07554559707641602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,1,power_law_1.01,0.09604480266571044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,1,power_law_1.01,0.09867519736289979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,1,power_law_1.01,0.10161279439926148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,1,power_law_1.01,0.10268800258636475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,1,power_law_1.01,0.1054527997970581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,1,power_law_1.01,0.10853760242462158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,1,power_law_1.01,0.1106112003326416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,1,power_law_1.01,0.11234560012817382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,1,power_law_1.01,0.1168895959854126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,1,power_law_1.01,0.1238976001739502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,1,power_law_1.01,0.12695679664611817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,1,power_law_1.01,0.1351423978805542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,1,power_law_1.01,0.15896960496902465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,1,power_law_1.01,0.18794239759445192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,1,power_law_1.01,0.23458559513092042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,1,power_law_1.01,0.2750080108642578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,1,power_law_1.01,0.3674623966217041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,1,power_law_1.01,0.45941758155822754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,1,power_law_1.01,0.6385087966918945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,1,power_law_1.01,0.8020799636840821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,1,power_law_1.01,1.1559679985046387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,1,balanced,5.732458750406901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,1,power_law_1.01,1.5050239562988281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,1,power_law_1.01,1.8649471282958985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,1,power_law_1.01,2.9193727493286135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,1,balanced,7.075482686360677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,1,power_law_1.01,5.763014221191407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,1,power_law_1.01,0.044300800561904906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,1,power_law_1.01,0.04657280147075653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,1,power_law_1.01,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,1,power_law_1.01,0.0638592004776001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,1,power_law_1.01,0.07244799733161926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,1,power_law_1.01,0.09363200068473816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,1,power_law_1.2,0.04467839896678925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,1,power_law_1.2,0.044736000895500186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,1,power_law_1.01,0.09525120258331299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,1,power_law_1.2,0.04924159944057464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,1,power_law_1.2,0.047328001260757445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,1,power_law_1.01,0.09749760031700135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,1,power_law_1.2,0.05625600218772888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,1,power_law_1.2,0.05124480128288269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,1,power_law_1.01,0.10106240510940552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,1,power_law_1.2,0.06410880088806152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,1,power_law_1.2,0.06259840130805969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,1,power_law_1.2,0.07351679801940918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,1,power_law_1.01,0.10215679407119752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,1,power_law_1.2,0.07093759775161743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,1,power_law_1.2,0.09715840220451355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,1,power_law_1.01,0.1072383999824524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,1,power_law_1.2,0.09387519955635071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,1,power_law_1.2,0.10243200063705445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,1,power_law_1.01,0.10661120414733886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,32,balanced,0.09118933478991191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,1,power_law_1.2,0.10064640045166015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,1,power_law_1.2,0.10371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,1,power_law_1.01,0.11176960468292237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,32,balanced,0.09332799911499023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,1,power_law_1.2,0.10439039468765259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,1,power_law_1.2,0.10179200172424316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,1,power_law_1.01,0.11623040437698365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,32,balanced,0.09168000022570293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,1,power_law_1.2,0.10117119550704956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,32,balanced,0.0997279981772105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,1,power_law_1.2,0.10691839456558228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,1,power_law_1.01,0.12444159984588624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,32,balanced,0.14805333813031515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,1,power_law_1.2,0.10426239967346192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,1,power_law_1.2,0.10915199518203736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,1,power_law_1.01,0.1317055940628052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,32,balanced,0.1641759971777598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,1,power_law_1.2,0.11080319881439209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,1,power_law_1.2,0.1057919979095459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,1,power_law_1.01,0.14269440174102782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,32,balanced,0.16354667147000632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,1,power_law_1.2,0.10885759592056274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,1,power_law_1.2,0.11388159990310669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,1,power_law_1.01,0.16906239986419677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,32,balanced,0.17081600427627563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,1,power_law_1.2,0.11344640254974366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,1,power_law_1.2,0.11782399415969849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,1,power_law_1.01,0.19912960529327392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,32,balanced,0.1723840037981669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,1,power_law_1.2,0.12414079904556274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,1,power_law_1.2,0.11800320148468017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,1,power_law_1.01,0.2502847909927368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,32,balanced,0.16797866423924765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,1,power_law_1.2,0.12562559843063353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,1,power_law_1.2,0.13089920282363893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,32,balanced,0.17384000619252524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,1,power_law_1.01,0.29949440956115725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,1,power_law_1.2,0.13355519771575927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,1,power_law_1.2,0.13880319595336915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,32,balanced,0.1762133240699768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,1,power_law_1.01,0.41157121658325196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,1,power_law_1.2,0.1624511957168579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,1,power_law_1.2,0.14309120178222656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,32,balanced,0.16107199589411417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,1,power_law_1.01,0.5234943866729737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,1,power_law_1.2,0.17342720031738282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,1,power_law_1.2,0.18955520391464234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,32,balanced,0.17162134250005087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,1,power_law_1.01,0.7436992168426514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,1,power_law_1.2,0.19921280145645143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,1,power_law_1.2,0.23630719184875487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,32,balanced,0.17402132352193198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,1,power_law_1.01,0.9544511795043945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,1,power_law_1.2,0.27336320877075193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,1,power_law_1.2,0.2534976005554199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,32,balanced,0.16219199697176614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,1,power_law_1.01,1.3933055877685547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,1,power_law_1.2,0.30110719203948977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,1,power_law_1.2,0.36805119514465334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,32,balanced,0.16790932416915894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,1,power_law_1.01,1.8228544235229491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,1,power_law_1.2,0.4121920108795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,1,power_law_1.2,0.45934081077575684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,1,power_law_1.2,0.5243840217590332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,1,power_law_1.2,0.6338047981262207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,32,balanced,0.16337600350379944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,1,power_law_1.01,2.263532829284668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,1,power_law_1.2,0.7430719852447509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,1,power_law_1.2,0.8040831565856934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,1,power_law_1.01,3.5740161895751954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,32,balanced,0.17326400677363077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,1,power_law_1.2,0.953439998626709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,1,power_law_1.2,1.1549568176269531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,1,power_law_1.2,1.3899968147277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,1,power_law_1.2,1.5029824256896973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,1,power_law_1.01,7.091865539550781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,32,balanced,0.1700213352839152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,1,power_law_1.2,1.820684814453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,1,power_law_1.2,1.8663999557495117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,1,power_law_1.2,2.2629888534545897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,1,power_law_1.2,2.9267200469970702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,32,balanced,0.17826666434605917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,1,power_law_1.2,3.5660480499267577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,1,power_law_1.2,5.768544006347656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,1,power_law_1.2,7.088896179199219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,32,balanced,0.26127467552820843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,32,balanced,0.24285866816838583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,32,balanced,0.3451146682103475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,32,balanced,0.4306826591491699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,32,balanced,0.4910879929860433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,32,balanced,0.6917706330617269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,32,balanced,0.7346346378326416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,32,balanced,0.040287998815377556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,32,balanced,0.04297066728274027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,32,balanced,1.1305867036183674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,32,balanced,0.0428959975639979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,32,balanced,0.06258666515350342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,32,balanced,0.042730664213498436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,32,balanced,0.06344000001748402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,32,balanced,0.04841599861780802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,32,balanced,0.06323199967543285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,32,balanced,0.07809066772460938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,32,balanced,0.059301331639289856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,32,balanced,0.0960746705532074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,32,balanced,0.0836853285630544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,32,balanced,0.10782933235168457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,32,balanced,0.08166400094827016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,32,balanced,0.10619733730951945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,32,balanced,0.07979199786980946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,32,balanced,0.11640000343322754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,32,balanced,0.09851732850074768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,32,balanced,0.11168000102043152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,32,balanced,0.09109866619110107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,32,balanced,0.1091306706269582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,32,balanced,0.0901759962240855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,32,balanced,0.11735999584197998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,32,balanced,0.0809440016746521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,32,balanced,0.11959999799728394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,32,balanced,0.0773173322280248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,32,balanced,0.11898666620254517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,32,balanced,0.10714133580525716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,32,balanced,0.11736533045768738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,32,balanced,0.10149866342544556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,32,balanced,0.11638399958610535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,32,balanced,0.12813333670298258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,32,balanced,0.08676266670227051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,32,balanced,0.12820266683896384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,32,balanced,0.10219200452168782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,32,balanced,0.1291253368059794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,32,balanced,0.12739200393358865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,32,balanced,0.13515200217564902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,32,balanced,0.16537066300710043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,32,balanced,0.14677332838376364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,32,balanced,0.2067199945449829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,32,balanced,0.1621226668357849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,32,balanced,0.2844640016555786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,32,balanced,0.2169599930445353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,32,balanced,0.3599253495534261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,32,balanced,0.23738133907318115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,32,balanced,0.5221120119094849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,32,balanced,0.3264426589012146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,32,balanced,0.6876160303751627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,32,balanced,2.215178648630778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,32,balanced,0.42546133200327557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,32,balanced,1.010538657506307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,32,balanced,0.5260693232218424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,32,balanced,1.3325440088907878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,32,balanced,0.7094506422678629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,32,balanced,1.647264003753662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,32,balanced,0.8097653388977051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,32,balanced,2.618874708811442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,32,balanced,1.2622666358947754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,32,balanced,0.0628959983587265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,32,balanced,0.062421331803003945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,32,balanced,0.06223999957243601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,32,balanced,0.07665599882602692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,32,balanced,0.09850133458773296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,32,balanced,0.1125333309173584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,32,balanced,0.10637866457303365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,32,balanced,0.11196266611417134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,32,balanced,0.10858666896820068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,32,balanced,0.11037333806355794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,32,balanced,0.10754133264223735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,32,balanced,0.10981333255767822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,32,balanced,0.11121066411336263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,32,balanced,0.10849600036938985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,32,balanced,0.12087466319402058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,32,balanced,0.11593066652615865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,32,balanced,0.12384532888730367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,32,balanced,0.12289067109425862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,32,balanced,0.12362133463223775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,32,balanced,5.193930625915527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,32,balanced,0.1288800040880839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,32,power_law_1.01,0.15986560583114623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,32,balanced,0.13521066308021545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,32,power_law_1.01,0.15324800014495848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,32,power_law_1.01,0.11357439756393432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,32,power_law_1.01,0.1155135989189148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,32,balanced,0.2050079902013143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,32,balanced,2.4834826787312827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,32,power_law_1.01,0.13337600231170654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,32,power_law_1.01,0.16022399663925171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,32,power_law_1.01,0.15674240589141847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,32,power_law_1.01,0.1538624048233032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,32,balanced,0.2100213368733724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,32,power_law_1.01,0.15794559717178344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,32,power_law_1.01,0.16173440217971802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,32,power_law_1.01,0.1578752040863037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,32,power_law_1.01,0.15469440221786498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,32,balanced,0.27568533023198444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,32,power_law_1.01,0.15957119464874267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,32,power_law_1.01,0.15746560096740722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,32,power_law_1.01,0.16374399662017822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,32,power_law_1.01,0.1595199942588806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,32,power_law_1.01,0.16180479526519775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,32,power_law_1.01,0.18462079763412476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,32,balanced,0.3502613306045532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,32,power_law_1.01,0.21767039299011232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,32,power_law_1.01,0.2531647920608521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,32,power_law_1.01,0.3064831972122192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,32,power_law_1.01,0.3560447931289673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,32,power_law_1.01,0.4401088237762451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,32,power_law_1.01,0.5400959968566894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,32,power_law_1.01,0.6899583816528321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,32,power_law_1.01,1.0699647903442382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,32,balanced,0.42071465651194256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,32,power_law_1.01,1.3737471580505372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,32,power_law_1.01,1.6752511978149414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,32,power_law_1.01,2.2124799728393554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,32,power_law_1.01,5.23864974975586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,32,balanced,0.5725440184275309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,32,balanced,0.6366453170776367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,32,balanced,0.9778133233388265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,32,power_law_1.01,0.11475839614868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,32,power_law_1.01,0.09858559966087341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,32,power_law_1.01,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,32,power_law_1.01,0.08328319787979126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,32,power_law_1.01,0.09658240079879761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,32,power_law_1.01,0.10784640312194824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,32,power_law_1.01,0.10435839891433715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,32,power_law_1.01,0.10970879793167114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,32,power_law_1.01,0.1093440055847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,32,power_law_1.01,0.10784640312194824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,32,power_law_1.01,0.10958720445632934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,32,power_law_1.01,0.11329280138015747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,32,power_law_1.01,0.10707839727401733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,32,power_law_1.01,0.11255680322647095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,32,power_law_1.01,0.10063999891281128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,32,power_law_1.01,0.11038719415664673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,32,power_law_1.01,0.11764479875564575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,32,power_law_1.01,0.07371519804000855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,32,power_law_1.01,0.11712000370025635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,32,power_law_1.01,0.08725119829177856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,32,power_law_1.01,0.12314239740371705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,32,power_law_1.01,0.08878719806671143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,32,power_law_1.01,0.1431040048599243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,32,power_law_1.01,0.10471680164337158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,32,power_law_1.01,0.15896320343017578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,32,power_law_1.01,0.10871039628982544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,32,power_law_1.01,0.19515520334243774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,32,power_law_1.01,0.10809600353240967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,32,power_law_1.01,0.2146752119064331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,32,power_law_1.01,0.10908160209655762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,32,power_law_1.01,0.10670080184936523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,32,power_law_1.01,0.28831360340118406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,32,power_law_1.01,0.10603519678115844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,32,power_law_1.01,0.3325376033782959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,32,power_law_1.01,0.10804480314254761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,32,power_law_1.01,0.482425594329834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,32,power_law_1.01,0.10147199630737305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,32,power_law_1.01,0.6143871784210205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,32,power_law_1.01,0.10801279544830322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,32,power_law_1.01,0.846457576751709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,32,power_law_1.01,0.11399680376052856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,32,power_law_1.01,1.103097629547119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,32,power_law_1.01,0.1088703989982605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,32,power_law_1.01,1.319206428527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,32,power_law_1.01,0.11527040004730224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,32,power_law_1.01,2.283001518249512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,32,power_law_1.01,0.13537280559539794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.01,0.147052800655365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,32,power_law_1.01,4.63191032409668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.01,0.17319040298461913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,32,balanced,1.9221173922220867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.01,0.1970687985420227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.01,0.26657919883728026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.01,0.28744959831237793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.01,0.4181248188018799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.01,0.5149759769439697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.01,0.6882751941680908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,32,power_law_1.01,0.03898240029811859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.01,0.9358719825744629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,32,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,32,power_law_1.01,1.098515224456787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,32,power_law_1.01,0.03495680093765259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,32,power_law_1.01,1.8430784225463868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,32,power_law_1.01,0.04068480134010315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,32,power_law_1.01,0.041510400176048276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,32,power_law_1.01,3.864710235595703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,32,power_law_1.01,0.05678719878196716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,32,power_law_1.01,0.0850816011428833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,32,power_law_1.01,0.08716800212860107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,32,power_law_1.01,0.0824832022190094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,32,power_law_1.01,0.09562879800796509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,32,power_law_1.01,0.09364479780197144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,32,power_law_1.01,0.09143679738044738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,32,power_law_1.01,0.08736000061035157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,32,power_law_1.01,0.08357759714126586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,32,power_law_1.01,0.08094080090522766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,32,power_law_1.01,0.08109440207481385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,32,power_law_1.01,0.07908480167388916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,32,power_law_1.01,0.08046079874038696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.01,0.08599680066108703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.01,0.11049599647521972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.01,0.12757760286331177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.01,0.16447360515594484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.01,0.20789120197296143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.01,0.28687999248504636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.01,0.34898560047149657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.01,0.5581503868103027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.01,0.651801586151123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,32,power_law_1.01,0.9263104438781739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,32,power_law_1.01,1.542188835144043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,32,power_law_1.01,2.701273536682129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,32,power_law_1.2,0.10042239427566528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,32,power_law_1.2,0.07400320172309875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,32,power_law_1.2,0.08047360181808472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,32,power_law_1.2,0.16078720092773438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,32,power_law_1.2,0.09082880020141601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,32,power_law_1.2,0.09129599928855896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,32,power_law_1.2,0.12743680477142333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,32,power_law_1.2,0.10099200010299683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,32,power_law_1.2,0.10561280250549317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,32,power_law_1.2,0.10641920566558838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,32,power_law_1.2,0.11327359676361085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,32,power_law_1.2,0.10943360328674316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,32,power_law_1.2,0.14743679761886597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,32,power_law_1.2,0.11400959491729737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,32,power_law_1.2,0.1693887948989868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,32,power_law_1.2,0.11004159450531006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,32,power_law_1.2,0.15627520084381102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,32,power_law_1.2,0.11066240072250366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,32,power_law_1.2,0.16349439620971679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,32,power_law_1.2,0.11340160369873047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,32,power_law_1.2,0.1621888041496277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,32,power_law_1.2,0.11140480041503906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,32,power_law_1.2,0.16039040088653564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,32,power_law_1.2,0.11244159936904907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,32,power_law_1.2,0.16105600595474243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,32,power_law_1.2,0.1189695954322815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,32,power_law_1.2,0.16614400148391723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,32,power_law_1.2,0.12223999500274658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,32,power_law_1.2,0.16106239557266236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,32,power_law_1.2,0.10586240291595458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,32,power_law_1.2,0.1250432014465332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,32,power_law_1.2,0.15927040576934814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,32,power_law_1.2,0.07954559922218322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,32,power_law_1.2,0.14782079458236694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,32,power_law_1.2,0.1622655987739563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,32,power_law_1.2,0.077292799949646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,32,power_law_1.2,0.16947840452194213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,32,power_law_1.2,0.17179520130157472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,32,power_law_1.2,0.0812608003616333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,32,power_law_1.2,0.21128320693969727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,32,power_law_1.2,0.18167680501937866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,32,power_law_1.2,0.08710399866104127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,32,power_law_1.2,0.24906880855560304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,32,power_law_1.2,0.20591359138488768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,32,power_law_1.2,0.09594240188598632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,32,power_law_1.2,0.33258240222930907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,32,power_law_1.2,0.22120320796966553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,32,power_law_1.2,0.10822399854660034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,32,power_law_1.2,0.41077117919921874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,32,power_law_1.2,0.29959681034088137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,32,power_law_1.2,0.1022271990776062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,32,power_law_1.2,0.5769087791442871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,32,power_law_1.2,0.30220799446105956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,32,power_law_1.2,0.11221120357513428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,32,power_law_1.2,0.7243072032928467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,32,power_law_1.2,0.42760958671569826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,32,power_law_1.2,0.10920319557189942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,32,power_law_1.2,0.5082304000854492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,32,power_law_1.2,1.1304320335388183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,32,power_law_1.2,0.10090240240097045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,32,power_law_1.2,0.7220928192138671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,32,power_law_1.2,1.5450559616088868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,32,power_law_1.2,0.11146240234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,32,power_law_1.2,1.0157888412475586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,32,power_law_1.2,1.5529919624328614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,32,power_law_1.2,0.10829440355300904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,32,power_law_1.2,1.428486442565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,32,power_law_1.2,3.0137727737426756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,32,power_law_1.2,0.11090559959411621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,32,power_law_1.2,2.224979209899902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,32,power_law_1.2,0.11809920072555542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,32,power_law_1.2,7.663078308105469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,32,power_law_1.2,2.289727973937988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,32,power_law_1.2,0.11898239850997924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,32,power_law_1.2,0.12198400497436523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,32,power_law_1.2,4.650809478759766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,32,power_law_1.2,0.14703999757766723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.2,0.15032320022583007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,32,power_law_1.2,8.949753570556641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.2,0.18869760036468505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.2,0.22144639492034912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.2,0.2857599973678589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.2,0.3409343957901001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.2,0.4919104099273682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.2,0.714739179611206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.2,1.0195648193359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,32,power_law_1.2,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,32,power_law_1.2,0.03309440016746521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.2,1.1772671699523927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,32,power_law_1.2,0.029600000381469725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,32,power_law_1.2,1.5682623863220215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,32,power_law_1.2,0.03866879940032959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,32,power_law_1.2,0.041715198755264284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,32,power_law_1.2,2.998918342590332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,32,power_law_1.2,0.06213120222091675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,32,power_law_1.2,5.406758499145508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,32,power_law_1.2,0.08350719809532166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,32,power_law_1.2,0.08652160167694092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,32,power_law_1.2,0.08485760092735291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,32,power_law_1.2,0.09464319944381713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,32,power_law_1.2,0.09164800047874451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,32,power_law_1.2,0.08701440095901489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,32,power_law_1.2,0.08923519849777221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,32,power_law_1.2,0.08069120049476623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,32,power_law_1.2,0.07921919822692872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,32,power_law_1.2,0.076883202791214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,32,power_law_1.2,0.07927039861679078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,32,power_law_1.2,0.07899519801139832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.2,0.09498239755630493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.2,0.11486079692840576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.2,0.14063359498977662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.2,0.18869119882583618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.2,0.24010879993438722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.2,0.3378688097000122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.2,0.44948477745056153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.2,0.7138879776000977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.2,0.8678976058959961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,32,power_law_1.2,1.196390438079834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,32,power_law_1.2,1.8809087753295899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,32,power_law_1.2,3.842937469482422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,16,balanced,0.06413333117961884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,16,balanced,0.06515199939409892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,16,balanced,0.07223999996980031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,16,balanced,0.097871998945872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,16,balanced,0.14959466457366943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,16,balanced,0.16346133748690286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,16,balanced,0.14473600188891092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,16,balanced,0.1557866632938385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,16,balanced,0.16076800227165222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,16,balanced,0.1479520003000895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,16,balanced,0.05449600021044413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,16,balanced,0.1516586641470591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,16,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,16,balanced,0.06214933097362518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,16,balanced,0.15794666608174643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,16,balanced,0.07379733522733052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,16,balanced,0.16592533389727274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,16,balanced,0.09877866506576538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,16,balanced,0.15191466609636942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,16,balanced,0.10659733414649963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,16,balanced,0.16641066471735635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,16,balanced,0.10800000031789143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,16,balanced,0.10698133707046509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,16,balanced,0.1569706698258718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,16,balanced,0.10955733060836792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,16,balanced,0.15916267037391663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,16,balanced,0.1036959985891978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,16,balanced,0.11052800218264262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,16,balanced,0.1607253352801005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,16,balanced,0.1063253382841746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,16,balanced,0.1639253298441569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,16,balanced,0.1071626643339793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,16,balanced,0.118559996287028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,16,balanced,0.17114667097727457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,16,balanced,0.11173866192499797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,16,balanced,0.05485333502292633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,16,balanced,0.055205335219701133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,16,balanced,0.11913067102432251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,16,balanced,0.06297066807746887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,16,balanced,0.12032000223795573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,16,balanced,0.18228266636530557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,16,balanced,0.07635733485221863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,16,balanced,0.09860799709955852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,16,balanced,0.12846933801968893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,16,balanced,0.11125333110491435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,16,balanced,0.10036266843477885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,16,balanced,0.13455999890963236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,16,balanced,0.26825066407521564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,16,balanced,0.10335466265678406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,16,balanced,0.11050132910410564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,16,balanced,0.1499733328819275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,16,balanced,0.10847466190656026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,16,balanced,0.10548800230026245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,16,balanced,0.24897066752115884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,16,balanced,0.10389332969983418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,16,balanced,0.16586132844289145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,16,balanced,0.10831999778747559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,16,balanced,0.1092693308989207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,16,balanced,0.11659733454386394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,16,balanced,0.3456159830093384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,16,balanced,0.22623467445373535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,16,balanced,0.11070400476455688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,16,balanced,0.1160533328851064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,16,balanced,0.12139200170834859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,16,balanced,0.2441813349723816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,16,balanced,0.12140267093976338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,16,balanced,0.4385066827138265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,16,balanced,0.1304853359858195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,16,balanced,0.141077329715093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,16,balanced,0.34059735139211017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,16,balanced,0.025727999707063038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,16,balanced,0.2012373407681783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,16,balanced,0.027061333258946735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,16,balanced,0.028309332827727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,16,balanced,0.5063999891281128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,16,balanced,0.02855466554562251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,16,balanced,0.430618683497111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,16,balanced,0.04008533308903376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,16,balanced,0.21241066853205362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,16,balanced,0.04923733572165171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,16,balanced,0.06229866544405619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,16,balanced,0.05888533095518748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,16,balanced,0.0689386675755183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,16,balanced,0.08382399876912434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,16,balanced,0.08444266517957051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,16,balanced,0.28830933570861816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,16,balanced,0.08747200171152751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,16,balanced,0.08730133374532063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,16,balanced,0.08614933490753174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,16,balanced,0.5577706495920817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,16,balanced,0.09981333216031392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,16,balanced,0.709216038386027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,16,balanced,0.09798933068911235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,16,balanced,0.09195733070373535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,16,balanced,0.37354131539662677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,16,balanced,0.10133866469065349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,16,balanced,0.12338133653004964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,16,balanced,0.14458133776982626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,16,balanced,0.16337066888809204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,16,balanced,0.7494293053944906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,16,balanced,0.20898133516311646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,16,balanced,0.45499734083811444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,16,balanced,0.7526400089263916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,16,balanced,0.2489173412322998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,16,balanced,0.3617173433303833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,16,balanced,0.866154670715332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,16,balanced,0.6088853279749552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,16,balanced,0.4483413298924764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,16,balanced,1.2001653512318928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,16,balanced,0.6525226831436157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,16,balanced,0.6925386587778727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,16,balanced,1.3523467381795247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,16,balanced,0.863813320795695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,16,balanced,1.0457706451416016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,16,balanced,1.0681119759877522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,16,balanced,2.3390132586161294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,16,balanced,1.6700480779012044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,16,balanced,2.6643306414286294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,16,balanced,2.0913119316101074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,16,balanced,3.2594292958577475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,16,power_law_1.01,0.13960319757461548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,16,power_law_1.01,0.14242559671401978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,16,power_law_1.01,0.08381440043449402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,16,power_law_1.01,0.10089600086212158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,16,power_law_1.01,0.1139456033706665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,16,power_law_1.01,0.14245760440826416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,16,power_law_1.01,0.14430079460144044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,16,power_law_1.01,0.1408511996269226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,16,power_law_1.01,0.14460159540176393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,16,power_law_1.01,0.14782079458236694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,16,power_law_1.01,0.09905920028686524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,16,power_law_1.01,0.14691200256347656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,16,power_law_1.01,0.09257599711418152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,16,power_law_1.01,0.1515071988105774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,16,power_law_1.01,0.06206079721450806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,16,power_law_1.01,0.1529855966567993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,16,power_law_1.01,0.07336959838867188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,16,power_law_1.01,0.1565824031829834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,16,power_law_1.01,0.09070720076560974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,16,power_law_1.01,0.1545215964317322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,16,power_law_1.01,0.10010240077972413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,16,power_law_1.01,0.10086400508880615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,16,power_law_1.01,0.15612159967422484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,16,power_law_1.01,0.10830719470977783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,16,power_law_1.01,0.1575808048248291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,16,power_law_1.01,0.10419199466705323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,16,power_law_1.01,0.1756608009338379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,16,power_law_1.01,0.10641920566558838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,16,power_law_1.01,0.19679360389709472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,16,power_law_1.01,0.10464639663696289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,16,power_law_1.01,0.21799681186676026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,16,power_law_1.01,0.1084928035736084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,16,power_law_1.01,0.26007680892944335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,16,power_law_1.01,0.11131520271301269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,16,power_law_1.01,0.3466048002243042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,16,power_law_1.01,0.10908160209655762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,16,power_law_1.01,0.382259202003479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,16,power_law_1.01,0.11633919477462769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,16,power_law_1.01,0.5177728176116944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,16,power_law_1.01,0.11644159555435181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,16,power_law_1.01,0.628275203704834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,16,power_law_1.01,0.12178560495376586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,16,power_law_1.01,0.8751296043395996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,16,power_law_1.01,0.09592959880828858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,16,power_law_1.01,0.13767679929733276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,16,power_law_1.01,1.2346495628356933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,16,power_law_1.01,0.09655680060386658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,16,power_law_1.01,0.15022720098495485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,16,power_law_1.01,0.06818559765815735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,16,power_law_1.01,1.3894335746765136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,16,power_law_1.01,0.18254079818725585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,16,power_law_1.01,0.07897599935531616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,16,power_law_1.01,0.21335039138793946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,16,power_law_1.01,2.023769569396973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,16,power_law_1.01,0.03273600041866302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,16,power_law_1.01,0.08417919874191285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,16,power_law_1.01,0.27314560413360595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,16,power_law_1.01,0.038700801134109494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,16,power_law_1.01,0.09898239970207215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,16,power_law_1.01,4.771583938598633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,16,power_law_1.01,0.3247423887252808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,16,power_law_1.01,0.02558720111846924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,16,power_law_1.01,0.10380159616470337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,16,power_law_1.01,0.4455999851226807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,16,power_law_1.01,0.027609598636627198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,16,power_law_1.01,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,16,power_law_1.01,0.5621119976043701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,16,power_law_1.01,0.02971520125865936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,16,power_law_1.01,0.10786559581756591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,16,power_law_1.01,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,16,power_law_1.01,0.8365119934082031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,16,power_law_1.01,0.10047999620437623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,16,power_law_1.01,0.057030397653579715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,16,power_law_1.01,1.0736063957214355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,16,power_law_1.01,0.1044160008430481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,16,power_law_1.01,0.05801600217819214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,16,power_law_1.01,0.10826239585876465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,16,power_law_1.01,1.3106240272521972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,16,power_law_1.01,0.0523967981338501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,16,power_law_1.01,0.10547840595245361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,16,power_law_1.01,1.9069887161254884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,16,power_law_1.01,0.10736000537872314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,16,power_law_1.01,0.08224639892578126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,16,power_law_1.01,0.08301439881324768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,16,power_law_1.01,0.11096320152282715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,16,power_law_1.01,4.129459381103516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,16,power_law_1.01,0.08382080197334289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,16,power_law_1.01,0.11439360380172729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,16,power_law_1.01,0.1187391996383667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,16,power_law_1.01,0.09566079974174499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,16,power_law_1.01,0.13452160358428955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,16,power_law_1.01,0.09123200178146362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.01,0.14311039447784424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,16,power_law_1.01,0.08390399813652039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.01,0.17086080312728882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,16,power_law_1.01,0.08305919766426087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.01,0.19413119554519653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,16,power_law_1.01,0.08354560136795045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,16,power_law_1.01,0.07860479950904846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.01,0.24495999813079833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.01,0.08263040184974671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.01,0.2858815908432007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.01,0.10123519897460938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.01,0.3976639986038208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.01,0.10995839834213257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.01,0.4768959999084473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.01,0.14243199825286865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.01,0.6810304164886475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.01,0.17209600210189818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.01,0.8082624435424804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.01,0.2364799976348877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,16,power_law_1.01,1.0180159568786622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.01,0.2744256019592285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,16,power_law_1.01,1.636511993408203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.01,0.41747198104858396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,16,power_law_1.01,3.934406280517578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.01,0.5572864055633545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,16,power_law_1.01,0.6477375984191894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,16,power_law_1.01,1.1113344192504884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,16,power_law_1.01,2.180505561828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,16,power_law_1.2,0.14654719829559326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,16,power_law_1.2,0.09105920195579528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,16,power_law_1.2,0.0868224024772644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,16,power_law_1.2,0.09160320162773132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,16,power_law_1.2,0.1081536054611206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,16,power_law_1.2,0.14319360256195068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,16,power_law_1.2,0.14211200475692748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,16,power_law_1.2,0.15006719827651976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,16,power_law_1.2,0.15267839431762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,16,power_law_1.2,0.15041279792785645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,16,power_law_1.2,0.15256960391998292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,16,power_law_1.2,0.15141760110855101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,16,power_law_1.2,0.1540287971496582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,16,power_law_1.2,0.15179519653320311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,16,power_law_1.2,0.160697603225708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,16,power_law_1.2,0.16543359756469728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,16,power_law_1.2,0.1652672052383423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,16,power_law_1.2,0.18807040452957152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,16,power_law_1.2,0.09864959716796876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,16,power_law_1.2,0.2041152000427246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,16,power_law_1.2,0.06625919938087463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,16,power_law_1.2,0.23928320407867432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,16,power_law_1.2,0.06135680079460144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,16,power_law_1.2,0.2886719942092896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,16,power_law_1.2,0.07393919825553893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,16,power_law_1.2,0.0828224003314972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,16,power_law_1.2,0.37690880298614504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,16,power_law_1.2,0.10126719474792481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,16,power_law_1.2,0.41770238876342775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,16,power_law_1.2,0.10258560180664063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,16,power_law_1.2,0.589683198928833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,16,power_law_1.2,0.10311679840087891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,16,power_law_1.2,0.6560319900512696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,16,power_law_1.2,0.10428160429000854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,16,power_law_1.2,1.072544002532959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,16,power_law_1.2,0.10671360492706299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,16,power_law_1.2,1.6927743911743165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,16,power_law_1.2,0.10878080129623413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,16,power_law_1.2,0.09757440090179444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,16,power_law_1.2,1.8480640411376954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,16,power_law_1.2,0.10951679944992065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,16,power_law_1.2,0.07279360294342041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,16,power_law_1.2,0.03185279965400696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,16,power_law_1.2,2.811334419250488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,16,power_law_1.2,0.1078719973564148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,16,power_law_1.2,0.06286079883575439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,16,power_law_1.2,0.0255295991897583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,16,power_law_1.2,0.1096127986907959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,16,power_law_1.2,0.07200639843940734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,16,power_law_1.2,6.304486465454102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,16,power_law_1.2,0.02470400035381317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,16,power_law_1.2,0.11586560010910034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,16,power_law_1.2,0.08362240195274354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,16,power_law_1.2,0.027059200406074523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,16,power_law_1.2,0.12315520048141479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,16,power_law_1.2,0.10175360441207885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,16,power_law_1.2,0.031052801012992858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,16,power_law_1.2,0.1265920042991638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,16,power_law_1.2,0.10055680274963379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,16,power_law_1.2,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,16,power_law_1.2,0.14646400213241578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,16,power_law_1.2,0.10572160482406616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,16,power_law_1.2,0.05742719769477844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,16,power_law_1.2,0.10185600519180298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,16,power_law_1.2,0.1598912000656128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,16,power_law_1.2,0.056460797786712646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,16,power_law_1.2,0.10435199737548828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,16,power_law_1.2,0.18910080194473267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,16,power_law_1.2,0.056435197591781616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,16,power_law_1.2,0.10892800092697144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,16,power_law_1.2,0.22148480415344238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,16,power_law_1.2,0.08284159898757934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,16,power_law_1.2,0.10661760568618775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,16,power_law_1.2,0.3104768037796021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,16,power_law_1.2,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,16,power_law_1.2,0.108787202835083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,16,power_law_1.2,0.34697599411010743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,16,power_law_1.2,0.0836736023426056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,16,power_law_1.2,0.11196800470352172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,16,power_law_1.2,0.5348927974700928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,16,power_law_1.2,0.09266560077667237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,16,power_law_1.2,0.11134079694747925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,16,power_law_1.2,0.6085951805114747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,16,power_law_1.2,0.08906239867210389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,16,power_law_1.2,0.11704319715499878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,16,power_law_1.2,0.08730880022048951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,16,power_law_1.2,0.9945023536682129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,16,power_law_1.2,0.12490880489349365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,16,power_law_1.2,0.08293759822845459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,16,power_law_1.2,1.2573439598083496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,16,power_law_1.2,0.1352512001991272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,16,power_law_1.2,0.08863999843597412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,16,power_law_1.2,1.6593215942382813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.2,0.15096319913864137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,16,power_law_1.2,0.08010240197181702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.2,0.17747199535369873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,16,power_law_1.2,2.4379199981689452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.2,0.08951039910316468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.2,0.10263680219650269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.2,0.21311359405517577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,16,power_law_1.2,4.750406265258789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.2,0.11939200162887573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.2,0.2663424015045166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.2,0.3398848056793213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.2,0.1500864028930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.2,0.1887168049812317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.2,0.47234559059143066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.2,0.25544960498809816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.2,0.5436736106872558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.2,0.31729280948638916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.2,0.9328255653381348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.2,0.4644735813140869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.2,0.9696703910827636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.2,0.620198392868042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,16,power_law_1.2,1.2084927558898926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,16,power_law_1.2,0.7491583824157715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,16,power_law_1.2,2.258393669128418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,16,power_law_1.2,1.2032832145690917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,16,power_law_1.2,4.504300689697265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,16,power_law_1.2,2.401900863647461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,8,balanced,0.0562720000743866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,8,balanced,0.06126399834950765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,8,balanced,0.072543998559316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,8,balanced,0.09398933251698811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,8,balanced,0.141893337170283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,8,balanced,0.15038399895032248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,8,balanced,0.15391467014948526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,8,balanced,0.14657066265741983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,8,balanced,0.150218665599823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,8,balanced,0.1460640033086141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,8,balanced,0.1481119990348816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,8,balanced,0.14522133270899454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,8,balanced,0.1476479967435201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,8,balanced,0.14883200327555338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,8,balanced,0.15481600165367126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,8,balanced,0.15461867054303488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,8,balanced,0.15321066975593567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,8,balanced,0.16645866632461548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,8,balanced,0.16760534048080444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,8,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,8,balanced,0.17844265699386597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,8,balanced,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,8,balanced,0.05625600119431814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,8,balanced,0.06897599995136261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,8,balanced,0.09784000118573506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,8,balanced,0.19049066305160522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,8,balanced,0.10613866647084554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,8,balanced,0.10477333267529805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,8,balanced,0.11053333679835002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,8,balanced,0.10883733630180359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,8,balanced,0.259226659933726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,8,balanced,0.10470400253931682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,8,balanced,0.1062506635983785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,8,balanced,0.10668800274531047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,8,balanced,0.2633333404858907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,8,balanced,0.10594666997591655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,8,balanced,0.11545600493748982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,8,balanced,0.052341332038243614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,8,balanced,0.11549333731333415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,8,balanced,0.05769066512584686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,8,balanced,0.06392000118891399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,8,balanced,0.11629333098729451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,8,balanced,0.024080000817775726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,8,balanced,0.3667999903361003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,8,balanced,0.08004799981911977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,8,balanced,0.025258667767047882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,8,balanced,0.12325867017110188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,8,balanced,0.10280000170071919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,8,balanced,0.02842666705449422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,8,balanced,0.11578133702278137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,8,balanced,0.1328480045000712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,8,balanced,0.02828266719977061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,8,balanced,0.11006933450698853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,8,balanced,0.04081599911053976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,8,balanced,0.11060266693433125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,8,balanced,0.14060266812642416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,8,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,8,balanced,0.11729600032170613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,8,balanced,0.05526400109132131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,8,balanced,0.1126026709874471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,8,balanced,0.46881067752838135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,8,balanced,0.051269332567850746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,8,balanced,0.16201600432395935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,8,balanced,0.11297600467999776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,8,balanced,0.05498133103052775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,8,balanced,0.12065066893895467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,8,balanced,0.06594133377075195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,8,balanced,0.11214400331179301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,8,balanced,0.1783519983291626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,8,balanced,0.0640533318122228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,8,balanced,0.11245333154996236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,8,balanced,0.06956799825032552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,8,balanced,0.11745599905649821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,8,balanced,0.0897226631641388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,8,balanced,0.11593066652615865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,8,balanced,0.23814932505289713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,8,balanced,0.08907199899355571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,8,balanced,0.12107200423876445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,8,balanced,0.5677760044733683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,8,balanced,0.08980799714724223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,8,balanced,0.13083199659983316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,8,balanced,0.09973866740862529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,8,balanced,0.2724799911181132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,8,balanced,0.1344586710135142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,8,balanced,0.10240532954533894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,8,balanced,0.11053333679835002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,8,balanced,0.14641599853833517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,8,balanced,0.12449066837628682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,8,balanced,0.37567468484242755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,8,balanced,0.16056533654530844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,8,balanced,0.13147200147310892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,8,balanced,0.22610666354497275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,8,balanced,0.7841066519419352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,8,balanced,0.1537920037905375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,8,balanced,0.4759253263473511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,8,balanced,0.2502346634864807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,8,balanced,0.18099733193715414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,8,balanced,0.33906133969624835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,8,balanced,0.20469866196314493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,8,balanced,0.6393599907557169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,8,balanced,0.4277919928232829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,8,balanced,0.8659466902414957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,8,balanced,0.3078186710675557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,8,balanced,0.8486773173014323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,8,balanced,0.5637653271357218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,8,balanced,0.3672959804534912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,8,balanced,0.7533493041992188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,8,balanced,1.3709653218587239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,8,balanced,1.0001386801401775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,8,balanced,0.879040002822876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,8,balanced,1.565951983133952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,8,balanced,1.3643520673116047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,8,balanced,2.6941280364990234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,8,balanced,3.069845199584961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,8,balanced,2.67087459564209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,8,balanced,0.5230506658554077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,8,balanced,0.6887199878692627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,8,power_law_1.01,0.09524480104446412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,8,power_law_1.01,0.10375679731369018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,8,power_law_1.01,0.0805184006690979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,8,power_law_1.01,0.09126399755477906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,8,power_law_1.01,0.10514559745788574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,8,power_law_1.01,0.13255679607391357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,8,balanced,0.8199893633524576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,8,power_law_1.01,0.141702401638031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,8,power_law_1.01,0.14402559995651246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,8,power_law_1.01,0.1410367965698242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,8,power_law_1.01,0.14540799856185913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,8,power_law_1.01,0.146943998336792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,8,power_law_1.01,0.15121279954910277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,8,power_law_1.01,0.14978560209274291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,8,power_law_1.01,0.1483456015586853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,8,power_law_1.01,0.15618560314178467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,8,power_law_1.01,0.15525120496749878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,8,power_law_1.01,0.15820159912109374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,8,power_law_1.01,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,8,power_law_1.01,0.17253119945526124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,8,power_law_1.01,0.08737279772758484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,8,power_law_1.01,0.1829568028450012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,8,power_law_1.01,0.062003201246261595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,8,power_law_1.01,0.22478721141815186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,8,power_law_1.01,0.0747327983379364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,8,power_law_1.01,0.24321279525756836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,8,power_law_1.01,0.0777728021144867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,8,power_law_1.01,0.3085439920425415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,8,power_law_1.01,0.09512959718704224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,8,power_law_1.01,0.3666815996170044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,8,power_law_1.01,0.09939839839935302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,8,power_law_1.01,0.5477759838104248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,8,power_law_1.01,0.10457600355148315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,8,balanced,1.3075253168741863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,8,power_law_1.01,0.6405439853668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,8,power_law_1.01,0.10179200172424316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,8,power_law_1.01,0.8122688293457031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,8,power_law_1.01,0.10533759593963624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,8,power_law_1.01,1.0329343795776367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,8,power_law_1.01,0.10357760190963745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,8,power_law_1.01,1.3550975799560547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,8,power_law_1.01,0.10822399854660034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,8,power_law_1.01,0.1084928035736084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,8,power_law_1.01,2.066092872619629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,8,power_law_1.01,0.07889919877052307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,8,power_law_1.01,0.11219840049743653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,8,power_law_1.01,3.7682174682617187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,8,power_law_1.01,0.09490560293197632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,8,power_law_1.01,0.11792000532150268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,8,power_law_1.01,0.06748160123825073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,8,power_law_1.01,0.12136319875717164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,8,power_law_1.01,0.0862335979938507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,8,power_law_1.01,0.12568960189819336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,8,power_law_1.01,0.08773760199546814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,8,power_law_1.01,0.13574399948120117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,8,power_law_1.01,0.110534405708313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,8,power_law_1.01,0.1488767981529236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,8,power_law_1.01,0.1078976035118103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,8,power_law_1.01,0.1830016016960144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,8,power_law_1.01,0.11471359729766846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,8,power_law_1.01,0.2142848014831543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,8,power_law_1.01,0.11127040386199952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,8,power_law_1.01,0.28799359798431395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,8,power_law_1.01,0.11025919914245605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,8,power_law_1.01,0.33664000034332275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,8,power_law_1.01,0.10995839834213257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,8,power_law_1.01,0.4483839988708496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,8,power_law_1.01,0.1146623969078064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,8,power_law_1.01,0.5691967964172363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,8,power_law_1.01,0.1165887951850891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,8,power_law_1.01,0.8476096153259277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,8,power_law_1.01,0.11616640090942383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,8,power_law_1.01,1.0353343963623047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,8,power_law_1.01,0.11937919855117798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,8,power_law_1.01,1.3780223846435546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,8,power_law_1.01,0.12275840044021606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,8,power_law_1.01,0.12970880270004273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,8,power_law_1.01,2.0727680206298826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,8,power_law_1.01,0.14088959693908693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,8,power_law_1.01,3.9173824310302736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.01,0.15449600219726561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.01,0.18605439662933348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.01,0.20292479991912843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.01,0.2603327989578247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.01,0.32727680206298826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.01,0.44284801483154296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.01,0.5457600116729736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.01,0.7493184089660645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,8,balanced,2.539013385772705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.01,0.8925632476806641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,8,power_law_1.01,1.1389504432678224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,8,power_law_1.01,1.8314239501953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,8,power_law_1.01,3.9716224670410156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,8,power_law_1.01,0.022918400168418885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,8,power_law_1.01,0.028787198662757873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,8,power_law_1.01,0.025158399343490602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,8,power_law_1.01,0.02632319927215576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,8,power_law_1.01,0.02942720055580139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,8,power_law_1.01,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,8,power_law_1.01,0.04359039962291718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,8,power_law_1.01,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,8,power_law_1.01,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,8,power_law_1.01,0.0577023983001709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,8,power_law_1.01,0.05667200088500977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,8,power_law_1.01,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,8,power_law_1.01,0.08312960267066956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,8,power_law_1.01,0.08423039913177491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,8,power_law_1.01,0.08432000279426574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,8,power_law_1.01,0.08846719861030579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,8,power_law_1.01,0.0911296010017395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,8,power_law_1.01,0.08899840116500854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,8,power_law_1.2,0.09457280039787293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.01,0.09023360013961793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,8,power_law_1.2,0.07174400091171265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.01,0.09845759868621826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,8,power_law_1.2,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.01,0.11065599918365479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,8,power_law_1.2,0.08006399869918823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.01,0.1324735999107361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,8,power_law_1.2,0.09719679951667785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.01,0.15839359760284424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,8,power_law_1.2,0.13162239789962768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.01,0.20146560668945312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,8,power_law_1.2,0.14153599739074707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.01,0.25365118980407714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,8,power_law_1.2,0.0707264006137848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,8,power_law_1.2,0.1430848002433777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.01,0.36250240802764894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,8,power_law_1.2,0.056377601623535153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,8,power_law_1.2,0.1452415943145752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.01,0.4952256202697754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,8,power_law_1.2,0.060844802856445314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,8,power_law_1.2,0.14886399507522582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,8,power_law_1.01,0.5780543804168701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,8,power_law_1.2,0.06859520077705383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,8,power_law_1.2,0.14701440334320068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,8,power_law_1.01,0.8722687721252441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,8,power_law_1.2,0.07690879702568054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,8,power_law_1.2,0.14828159809112548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,8,power_law_1.2,0.08916479945182801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,8,power_law_1.2,0.15003520250320435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,8,power_law_1.01,1.7732927322387695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,8,power_law_1.2,0.10024319887161255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,8,power_law_1.2,0.14822399616241455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,8,power_law_1.2,0.10389120578765869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,8,power_law_1.2,0.15845760107040405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,8,power_law_1.2,0.10229760408401489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,8,power_law_1.2,0.16154240369796752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,8,power_law_1.2,0.1032639980316162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,8,power_law_1.2,0.16591999530792237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,8,power_law_1.2,0.07952640056610108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,8,power_law_1.2,0.1049407958984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,8,power_law_1.2,0.18039679527282715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,8,power_law_1.2,0.06058239936828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,8,power_law_1.2,0.1076416015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,8,power_law_1.2,0.1912832021713257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,8,power_law_1.2,0.06480640172958374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,8,power_law_1.2,0.10680960416793824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,8,power_law_1.2,0.23374719619750978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,8,power_law_1.2,0.07791360020637512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,8,power_law_1.2,0.11278079748153687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,8,power_law_1.2,0.08584960103034973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,8,power_law_1.2,0.2655103921890259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,8,power_law_1.2,0.121343994140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,8,power_law_1.2,0.10381439924240113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,8,power_law_1.2,0.3366208076477051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,8,power_law_1.2,0.12201600074768067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,8,power_law_1.2,0.10288000106811523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,8,power_law_1.2,0.36672000885009765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,8,power_law_1.2,0.12604160308837892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,8,power_law_1.2,0.11225600242614746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,8,power_law_1.2,0.5618368148803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,8,power_law_1.2,0.1461184024810791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,8,power_law_1.2,0.10964479446411132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,8,power_law_1.2,0.637824010848999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,8,power_law_1.2,0.15440640449523926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,8,power_law_1.2,0.11366399526596069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,8,power_law_1.2,0.9768959999084472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,8,power_law_1.2,0.19733760356903077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,8,power_law_1.2,0.11017600297927857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,8,power_law_1.2,1.283353614807129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,8,power_law_1.2,0.2259135961532593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,8,power_law_1.2,0.11357439756393432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,8,power_law_1.2,1.4970623970031738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,8,power_law_1.2,0.12028160095214843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,8,power_law_1.2,0.28551039695739744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,8,power_law_1.2,0.11856640577316284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,8,power_law_1.2,0.3701632022857666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,8,power_law_1.2,2.7033151626586913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,8,power_law_1.2,0.12429440021514893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,8,power_law_1.2,0.4965184211730957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,8,power_law_1.2,0.1265663981437683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,8,power_law_1.2,5.573875045776367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,8,power_law_1.2,0.6239615917205811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,8,power_law_1.2,0.1295232057571411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,8,power_law_1.2,0.8845312118530273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,8,power_law_1.2,0.14543999433517457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,8,power_law_1.2,1.1565247535705567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.2,0.16350079774856568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,8,power_law_1.2,1.7844863891601563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.2,0.18380160331726075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,8,power_law_1.2,2.46374397277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.2,0.22104320526123047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.2,0.28606081008911133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,8,power_law_1.2,4.374841690063477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.2,0.35598719120025635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.2,0.5230527877807617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.2,0.5876095771789551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.2,0.8386431694030761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.2,1.2070783615112304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,8,power_law_1.2,1.3867648124694825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,8,power_law_1.2,2.297043228149414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,8,power_law_1.2,5.31767692565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,8,power_law_1.2,0.02369920015335083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,8,power_law_1.2,0.023174400627613067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,8,power_law_1.2,0.025433599948883057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,8,power_law_1.2,0.02656640112400055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,8,power_law_1.2,0.02810240089893341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,8,power_law_1.2,0.03802880048751831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,8,power_law_1.2,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,8,power_law_1.2,0.048467200994491574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,8,power_law_1.2,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,8,power_law_1.2,0.05674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,8,power_law_1.2,0.056364798545837404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,8,power_law_1.2,0.05676159858703613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,8,power_law_1.2,0.08279039859771728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,8,power_law_1.2,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,8,power_law_1.2,0.08465920090675354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,8,power_law_1.2,0.09390720129013061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,8,power_law_1.2,0.09126399755477906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,8,power_law_1.2,0.085452800989151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.2,0.09169279932975768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.2,0.10665600299835205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.2,0.11146880388259887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.2,0.13954559564590455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.2,0.1672767996788025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.2,0.22992639541625975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,4,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.2,0.2861056089401245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,4,balanced,0.057002668579419456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,4,balanced,0.058506667613983154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,4,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.2,0.37202560901641846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,4,balanced,0.06356800099213918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,4,balanced,0.07527466615041097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,4,balanced,0.07578666508197784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.2,0.5125120162963868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,4,balanced,0.10981333255767822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,4,balanced,0.09501333038012187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,4,balanced,0.1184266706307729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,8,power_law_1.2,0.6234240055084228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,4,balanced,0.14448533455530801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,4,balanced,0.11622933546702068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,4,balanced,0.16275200247764587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,8,power_law_1.2,1.0252927780151366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,4,balanced,0.1167093316713969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,4,balanced,0.14500266313552856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,4,balanced,0.11766933401425679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,4,balanced,0.14866666992505392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,8,power_law_1.2,2.003321647644043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,4,balanced,0.11611200372378032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,4,balanced,0.14617600043614706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,4,balanced,0.11798933148384094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,4,balanced,0.14587199687957764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,4,balanced,0.11867733796437581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,4,balanced,0.14643733700116476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,4,balanced,0.11858133474985759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,4,balanced,0.14735466241836548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,4,balanced,0.12150933345158894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,4,balanced,0.14753599961598715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,4,balanced,0.12909866372744241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,4,balanced,0.14894400040308634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,4,balanced,0.1306613286336263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,4,balanced,0.1556426684061686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,4,balanced,0.1388320028781891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,4,balanced,0.1557866632938385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,4,balanced,0.15201066931088766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,4,balanced,0.16208533445994058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,4,power_law_1.01,0.07466880083084107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,4,balanced,0.16211199760437012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,4,power_law_1.01,0.0877951979637146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,4,balanced,0.1742560068766276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,4,power_law_1.01,0.07776640057563781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,4,balanced,0.1869600017865499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,4,power_law_1.01,0.0897599995136261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,4,balanced,0.18062933286031088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,4,power_law_1.01,0.09964159727096558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,4,power_law_1.01,0.13914239406585693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,4,balanced,0.19989866018295288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,4,balanced,0.21273066600163779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,4,power_law_1.01,0.14406399726867675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,4,power_law_1.01,0.14373120069503784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,4,balanced,0.28990399837493896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,4,balanced,0.21817066272099814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,4,power_law_1.01,0.14484479427337646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,4,power_law_1.01,0.14873600006103516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,4,power_law_1.01,0.14940799474716188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,4,balanced,0.3374933401743571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,4,power_law_1.01,0.14489599466323852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,4,balanced,0.2896266579627991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,4,power_law_1.01,0.15146880149841307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,4,power_law_1.01,0.15036799907684326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,4,power_law_1.01,0.1610751986503601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,4,balanced,0.4697813193003337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,4,power_law_1.01,0.16121599674224854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,4,balanced,0.3219520052274068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,4,power_law_1.01,0.16657279729843139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,4,power_law_1.01,0.17968000173568727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,4,power_law_1.01,0.1912384033203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,4,balanced,0.6052053372065226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,4,power_law_1.01,0.22847359180450438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,4,power_law_1.01,0.25343999862670896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,4,balanced,0.4312533140182495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,4,power_law_1.01,0.31415040493011476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,4,power_law_1.01,0.3701567888259888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,4,power_law_1.01,0.5067584037780761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,4,power_law_1.01,0.617958402633667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,4,power_law_1.01,0.8649151802062989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,4,balanced,0.8261813322703043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,4,power_law_1.01,1.0921024322509765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,4,power_law_1.01,1.412928009033203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,4,balanced,0.5668960014979044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,4,power_law_1.01,1.9248832702636718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,4,power_law_1.01,4.042950439453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,4,balanced,1.0912000338236492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,4,balanced,0.7178773085276285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,4,balanced,1.302085320154826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,4,balanced,0.9535733064015707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,4,balanced,2.051402727762858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,4,balanced,1.1020639737447102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,4,power_law_1.01,0.06460800170898437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,4,power_law_1.01,0.07411199808120728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,4,power_law_1.01,0.06299520134925843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,4,power_law_1.01,0.07309439778327942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,4,power_law_1.01,0.08575360178947448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,4,power_law_1.01,0.1003775954246521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,4,power_law_1.01,0.1118016004562378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,4,power_law_1.01,0.11194239854812622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,4,power_law_1.01,0.1149440050125122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,4,power_law_1.01,0.11628799438476563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,4,power_law_1.01,0.11975040435791015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,4,power_law_1.01,0.11795200109481811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,4,power_law_1.01,0.11989120244979859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,4,power_law_1.01,0.12430720329284668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,4,power_law_1.01,0.1306175947189331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,4,power_law_1.01,0.13542399406433106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,4,power_law_1.01,0.14055039882659912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,4,power_law_1.01,0.15564800500869752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,4,power_law_1.01,0.17489919662475586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,4,power_law_1.01,0.2174527883529663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,4,balanced,1.7307413419087727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,4,power_law_1.01,0.24428160190582277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,4,power_law_1.01,0.3169536113739014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,4,power_law_1.01,0.382425594329834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,4,power_law_1.01,0.524019193649292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,4,power_law_1.01,0.6987071990966797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,4,balanced,4.043087959289551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,4,power_law_1.01,0.9670463562011719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,4,power_law_1.01,1.2509568214416504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,4,power_law_1.01,1.5528639793395995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,4,power_law_1.01,2.4067583084106445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,4,power_law_1.01,4.632134246826172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,4,power_law_1.2,0.07482879757881164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,4,power_law_1.2,0.07192959785461425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,4,power_law_1.2,0.0719488024711609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,4,power_law_1.2,0.08481280207633972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,4,power_law_1.2,0.09681280255317688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,4,power_law_1.2,0.1333631992340088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,4,power_law_1.2,0.14479999542236327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,4,power_law_1.2,0.1466752052307129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,4,power_law_1.2,0.1455296039581299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,4,power_law_1.2,0.1459391951560974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,4,power_law_1.2,0.1501952052116394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,4,power_law_1.2,0.15292799472808838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,4,power_law_1.2,0.151910400390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,4,power_law_1.2,0.14833279848098754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,4,power_law_1.2,0.16302720308303834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,4,power_law_1.2,0.16517119407653807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,4,power_law_1.2,0.16990079879760742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,4,power_law_1.2,0.18318079710006713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,4,power_law_1.2,0.19552639722824097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,4,power_law_1.2,0.2412480115890503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,4,power_law_1.2,0.26987519264221194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,4,power_law_1.2,0.3252543926239014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,4,power_law_1.2,0.4197887897491455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,4,balanced,3.4170347849527993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,4,power_law_1.2,0.5310719966888428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,4,power_law_1.2,0.6274303913116455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,4,power_law_1.2,0.9711615562438964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,4,power_law_1.2,1.2512639999389648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,4,power_law_1.2,1.6087039947509765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,4,power_law_1.2,2.44006404876709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,4,power_law_1.2,4.875699234008789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,4,power_law_1.2,0.06350079774856568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,4,power_law_1.2,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,4,power_law_1.2,0.06219519972801209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,4,power_law_1.2,0.07382400035858154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,4,power_law_1.2,0.08183680176734924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,4,power_law_1.2,0.09791359901428223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,4,power_law_1.2,0.11034879684448243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,4,power_law_1.2,0.11322879791259766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,2,balanced,0.06247466802597046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,2,balanced,0.06779733300209045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,4,power_law_1.2,0.1136512041091919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,2,balanced,0.08331733445326488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,4,power_law_1.2,0.11496959924697876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,2,balanced,0.10574932893117268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,4,power_law_1.2,0.11764479875564575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,2,balanced,0.1670773426691691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,2,balanced,0.17728533347447714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,4,power_law_1.2,0.12163200378417968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,2,balanced,0.17078399658203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,4,power_law_1.2,0.12202240228652954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,2,balanced,0.17247466246287027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,4,power_law_1.2,0.12463359832763672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,2,balanced,0.16680532693862915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,4,power_law_1.2,0.13219840526580812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,2,balanced,0.17154133319854736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,4,power_law_1.2,0.13736319541931152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,2,balanced,0.170415997505188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,4,power_law_1.2,0.14395519495010375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,2,balanced,0.17226133743921915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,4,power_law_1.2,0.15857919454574584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,2,balanced,0.1723573406537374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,4,power_law_1.2,0.17441279888153077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,2,balanced,0.17315733432769775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,4,power_law_1.2,0.21901440620422363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,2,balanced,0.17985600233078003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,4,power_law_1.2,0.25372800827026365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,2,balanced,0.18018666903177896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,4,power_law_1.2,0.3189183950424194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,2,balanced,0.18836800257364908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,4,power_law_1.2,0.37719039916992186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,2,balanced,0.2084746758143107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,4,power_law_1.2,0.6072127819061279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,2,balanced,0.22057066361109415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,4,power_law_1.2,0.7362815856933593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,4,power_law_1.2,1.0561792373657226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,2,balanced,0.2531893253326416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,4,power_law_1.2,1.4795904159545898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,4,power_law_1.2,1.676153564453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,2,balanced,0.2823253273963928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,4,power_law_1.2,2.8088512420654297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,2,balanced,0.4129439989725749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,4,power_law_1.2,5.747097778320312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,2,balanced,0.46860798199971515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,2,balanced,0.6493706703186035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,2,balanced,0.056847999493281044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,2,balanced,0.06108800073464712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,2,balanced,0.07156800230344136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,2,balanced,0.0899679958820343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,2,balanced,0.13218667109807333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,2,balanced,0.8361600240071615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,2,balanced,0.1415893336137136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,2,balanced,0.142277330160141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,2,balanced,0.14538133144378662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,2,balanced,0.1437066694100698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,2,balanced,0.14176000157992044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,2,balanced,0.1418239971001943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,2,balanced,0.14543466766675314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,2,balanced,0.14425599575042725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,2,power_law_1.01,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,2,balanced,0.15101333459218344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,2,balanced,1.0985759894053142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,2,power_law_1.01,0.07919359803199769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,2,balanced,0.15544533729553223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,2,power_law_1.01,0.07716479897499084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,2,balanced,0.16010133425394693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,2,power_law_1.01,0.09596800208091735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,2,balanced,0.16771199305852255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,2,power_law_1.01,0.11366399526596069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,2,balanced,0.18819733460744223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,2,power_law_1.01,0.1549183964729309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,2,power_law_1.01,0.1642624020576477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,2,balanced,0.20302400986353555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,2,power_law_1.01,0.1591104030609131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,2,power_law_1.01,0.16279040575027465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,2,balanced,0.24207999308904013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,2,balanced,1.4767573674519856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,2,power_law_1.01,0.16338560581207276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,2,balanced,0.27478400866190594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,2,power_law_1.01,0.16493439674377441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,2,power_law_1.01,0.16787199974060057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,2,power_law_1.01,0.1692415952682495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,2,balanced,0.40673065185546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,2,power_law_1.01,0.17653759717941284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,2,power_law_1.01,0.18417919874191285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,2,power_law_1.01,0.18597760200500488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,2,power_law_1.01,0.1935744047164917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,2,balanced,0.47811734676361084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,2,power_law_1.01,0.2157439947128296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,2,power_law_1.01,0.23493759632110595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,2,power_law_1.01,0.287225604057312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,2,balanced,1.7491040229797363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,2,power_law_1.01,0.33536639213562014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,2,balanced,0.6718346277872721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,2,power_law_1.01,0.42892799377441404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,2,power_law_1.01,0.5100863933563232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,2,power_law_1.01,0.7051839828491211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,2,power_law_1.01,0.9057344436645508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,2,power_law_1.01,1.1903488159179687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,2,balanced,0.8751733303070068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,2,power_law_1.01,1.5689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,2,power_law_1.01,1.9639616012573242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,2,power_law_1.01,3.1372032165527344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,2,power_law_1.01,6.079763031005859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,2,balanced,1.2070186932881672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,2,balanced,2.771162668863932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,2,balanced,1.5990293820699055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,2,balanced,1.9352854092915852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,2,power_law_1.01,0.06218240261077881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,2,power_law_1.01,0.06693120002746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,2,power_law_1.01,0.07055360078811646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,2,power_law_1.01,0.08386560082435608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,2,power_law_1.01,0.09608960151672363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,2,power_law_1.01,0.12518399953842163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,2,power_law_1.01,0.13241599798202514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,2,balanced,5.504506429036458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,2,power_law_1.01,0.1332927942276001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,2,power_law_1.01,0.13171199560165406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,2,power_law_1.01,0.13645440340042114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,2,power_law_1.01,0.13317760229110717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,2,balanced,3.0609280268351235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,2,power_law_1.01,0.13852800130844117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,2,power_law_1.01,0.14266239404678344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,2,power_law_1.01,0.14520959854125975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,2,power_law_1.01,0.15459200143814086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,2,power_law_1.01,0.16021759510040284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,2,power_law_1.01,0.17108479738235474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,2,power_law_1.01,0.19191039800643922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,2,power_law_1.01,0.21644160747528077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,2,power_law_1.01,0.2702143907546997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,2,power_law_1.01,0.31975679397583007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,2,power_law_1.01,0.42259202003479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,2,power_law_1.01,0.5384128093719482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,2,power_law_1.01,0.6917312145233154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,2,power_law_1.01,0.904633617401123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,2,power_law_1.01,1.258188819885254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,2,power_law_1.01,1.7228864669799804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,2,power_law_1.01,2.1591487884521485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,2,power_law_1.01,3.3611968994140624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,2,power_law_1.01,6.836844635009766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,2,balanced,6.068458557128906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,2,power_law_1.2,0.067603200674057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,2,power_law_1.2,0.07036160230636597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,2,power_law_1.2,0.07557759881019592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,2,power_law_1.2,0.09281280040740966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,2,power_law_1.2,0.1079103946685791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,2,power_law_1.2,0.1532096028327942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,2,power_law_1.2,0.15953919887542725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,2,power_law_1.2,0.16563839912414552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,2,power_law_1.2,0.165011203289032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,2,power_law_1.2,0.1664896011352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,2,power_law_1.2,0.1681663990020752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,2,power_law_1.2,0.16960639953613282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,2,power_law_1.2,0.17323520183563232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,2,power_law_1.2,0.17656960487365722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,2,power_law_1.2,0.18487679958343506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,2,power_law_1.2,0.18863359689712525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,2,power_law_1.2,0.19949439764022828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,2,power_law_1.2,0.22289280891418456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,2,power_law_1.2,0.2511039972305298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,2,power_law_1.2,0.29243519306182864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,2,power_law_1.2,0.33987200260162354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,2,power_law_1.2,0.43927679061889646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,2,power_law_1.2,0.5480832099914551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,2,power_law_1.2,0.7138432025909424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,2,power_law_1.2,0.9302463531494141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,2,power_law_1.2,1.327564811706543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,2,power_law_1.2,1.6992895126342773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,2,power_law_1.2,2.1259584426879883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,2,power_law_1.2,3.407148742675781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,2,power_law_1.2,6.377721786499023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,2,power_law_1.2,0.06254720091819763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,2,power_law_1.2,0.06173440217971802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,2,power_law_1.2,0.06764159798622131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,2,power_law_1.2,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,2,power_law_1.2,0.09205120205879211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,2,power_law_1.2,0.12435200214385986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,2,power_law_1.2,0.13318400382995604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,2,power_law_1.2,0.13282560110092162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,2,power_law_1.2,0.13941119909286498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,2,power_law_1.2,0.13659520149230958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,2,power_law_1.2,0.13857280015945433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,2,power_law_1.2,0.14343039989471434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,2,power_law_1.2,0.14385279417037963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,2,power_law_1.2,0.14780800342559813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,2,power_law_1.2,0.1561087965965271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,2,power_law_1.2,0.15964159965515137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,2,power_law_1.2,0.1736256003379822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,2,power_law_1.2,0.1956928014755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,2,power_law_1.2,0.21434879302978516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,2,power_law_1.2,0.27550721168518066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,1,balanced,0.06914666791756947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,2,power_law_1.2,0.32622718811035156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,1,balanced,0.07894933223724365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,1,balanced,0.0992693305015564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,2,power_law_1.2,0.42992639541625977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,1,balanced,0.13897599776585898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,2,power_law_1.2,0.525600004196167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,1,balanced,0.2067413330078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,1,balanced,0.2524320085843404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,2,power_law_1.2,0.770201587677002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,1,balanced,0.2437173326810201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,2,power_law_1.2,0.9431936264038085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,1,balanced,0.24279467264811197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,2,power_law_1.2,1.4902976036071778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,1,balanced,0.2403093377749125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,1,balanced,0.24155733982721964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,2,power_law_1.2,1.8611328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,1,balanced,0.2428426742553711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,2,power_law_1.2,2.2705408096313477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,1,balanced,0.24181866645812988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,2,power_law_1.2,3.6069568634033202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,1,balanced,0.2421333392461141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,1,balanced,0.2474773327509562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,2,power_law_1.2,7.56033935546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,1,balanced,0.254586656888326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,1,balanced,0.2632426619529724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,1,balanced,0.266159991423289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,1,balanced,0.2988160053888957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,1,balanced,0.3212266763051351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,1,balanced,0.3783786694208781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,1,balanced,0.06355200211207072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,1,balanced,0.07264000177383423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,1,balanced,0.08290133376916249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,1,balanced,0.428117314974467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,1,balanced,0.11428266763687134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,1,balanced,0.16607999801635742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,1,balanced,0.19379733006159464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,1,balanced,0.1947093407313029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,1,balanced,0.6380639870961508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,1,balanced,0.19747199614842734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,1,balanced,0.2000160018603007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,1,balanced,0.20290666818618774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,1,balanced,0.20619734128316244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,1,balanced,0.7271626790364584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,1,balanced,0.2067413330078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,1,balanced,0.20880534251530966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,1,balanced,0.21701866388320923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,1,balanced,0.22143999735514322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,1,balanced,0.22797866662343344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,1,balanced,1.0329173405965169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,1,balanced,0.2402986685434977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,1,balanced,0.27772800127665204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,1,balanced,0.3010080059369405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,1,balanced,0.3670453230539958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,1,balanced,1.3433653513590496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,1,balanced,0.4235359827677409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,1,balanced,0.6352373361587524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,1,balanced,0.7485333283742269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,1,balanced,1.8350879351298015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,1,balanced,1.088762680689494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,1,balanced,1.4216853777567546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,1,balanced,2.4468159675598145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,1,power_law_1.01,0.0662015974521637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,1,power_law_1.01,0.07599999904632568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,1,power_law_1.01,0.0892799973487854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,1,power_law_1.01,0.11176960468292237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,1,balanced,1.9734773635864258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,1,power_law_1.01,0.1315392017364502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,1,power_law_1.01,0.18804479837417604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,1,power_law_1.01,0.19315199851989745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,1,power_law_1.01,0.20611839294433593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,1,power_law_1.01,0.20419199466705323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,1,power_law_1.01,0.21020159721374512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,1,power_law_1.01,0.213702392578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,1,power_law_1.01,0.21475839614868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,1,power_law_1.01,0.22198400497436524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,1,balanced,2.9393386840820312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,1,power_law_1.01,0.22951679229736327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,1,power_law_1.01,0.23605120182037354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,1,balanced,2.623136043548584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,1,power_law_1.01,0.24316799640655518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,1,power_law_1.01,0.2619712114334106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,1,power_law_1.01,0.3019200086593628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,1,power_law_1.01,0.3337023973464966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,1,power_law_1.01,0.3995136022567749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,1,power_law_1.01,0.4773375988006592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,1,power_law_1.01,0.6198527812957764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,1,power_law_1.01,0.7482240200042725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,1,power_law_1.01,1.0341567993164062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,1,power_law_1.01,1.3179583549499512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,1,power_law_1.01,1.8965951919555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,1,power_law_1.01,2.4820032119750977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,1,balanced,3.2070134480794272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,1,power_law_1.01,3.054547119140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,1,power_law_1.01,4.808441543579102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,1,power_law_1.01,0.06250879764556885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,1,power_law_1.01,0.06813439726829529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,1,power_law_1.01,9.61690902709961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,1,power_law_1.01,0.07825919985771179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,1,balanced,4.689925193786621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,1,power_law_1.01,0.10343040227890014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,1,power_law_1.01,0.11444480419158935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,1,power_law_1.01,0.15481599569320678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,1,power_law_1.01,0.16631040573120118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,1,power_law_1.01,0.16993919610977173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,1,power_law_1.01,0.17522560358047484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,1,power_law_1.01,0.17594879865646362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,1,power_law_1.01,0.18315520286560058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,1,power_law_1.01,0.1919935941696167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,1,power_law_1.01,0.18823039531707764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,1,power_law_1.01,0.19927040338516236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,1,power_law_1.01,0.20952320098876953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,1,balanced,5.110927899678548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,1,power_law_1.01,0.21662719249725343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,1,power_law_1.01,0.23683199882507325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,1,power_law_1.01,0.27589759826660154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,1,power_law_1.01,0.3099008083343506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,1,power_law_1.01,0.38765439987182615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,1,power_law_1.01,0.46250238418579104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,1,power_law_1.01,0.6207104206085206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,1,power_law_1.01,0.769593620300293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,1,power_law_1.01,1.0766719818115233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,1,power_law_1.01,1.3890111923217774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,1,power_law_1.01,2.00695686340332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,1,power_law_1.01,2.620275115966797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,1,power_law_1.01,3.265248107910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,1,power_law_1.01,5.1443328857421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,1,power_law_1.01,10.251795196533203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,1,balanced,9.340346654256185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,1,balanced,10.112821578979492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,1,power_law_1.2,0.0667136013507843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,1,power_law_1.2,0.07432960271835327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,1,power_law_1.2,0.08651520013809204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,1,power_law_1.2,0.10877439975738526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,1,power_law_1.2,0.12464640140533448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,1,power_law_1.2,0.18888319730758668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,1,power_law_1.2,0.19533439874649047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,1,power_law_1.2,0.20865280628204347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,1,power_law_1.2,0.20885119438171387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,1,power_law_1.2,0.21461760997772217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,1,power_law_1.2,0.21566081047058105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,1,power_law_1.2,0.2271807909011841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,1,power_law_1.2,0.23623039722442626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,1,power_law_1.2,0.23797760009765626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,1,power_law_1.2,0.2530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,1,power_law_1.2,0.2591615915298462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,1,power_law_1.2,0.2705984115600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,1,power_law_1.2,0.30796160697937014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,1,power_law_1.2,0.33999359607696533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,1,power_law_1.2,0.40761599540710447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,1,power_law_1.2,0.06049280166625977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,1,power_law_1.2,0.48334717750549316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,1,power_law_1.2,0.06425600051879883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,1,power_law_1.2,0.630675220489502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,1,power_law_1.2,0.07941120266914367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,1,power_law_1.2,0.7560128211975098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,1,power_law_1.2,0.09388800263404846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,1,power_law_1.2,1.0493887901306151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,1,power_law_1.2,0.10781439542770385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,1,power_law_1.2,0.15489280223846436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,1,power_law_1.2,1.3369407653808594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,1,power_law_1.2,0.16403839588165284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,1,power_law_1.2,1.9147008895874023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,1,power_law_1.2,0.16811519861221313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,1,power_law_1.2,2.509388732910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,1,power_law_1.2,0.17627520561218263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,1,power_law_1.2,3.10327033996582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,1,power_law_1.2,0.1799999952316284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,1,power_law_1.2,0.18595839738845826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,1,power_law_1.2,4.850604629516601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,1,power_law_1.2,0.18946559429168702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,1,power_law_1.2,0.19596799612045288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,1,power_law_1.2,9.597151947021484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,32,balanced,0.08794132868448894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,1,power_law_1.2,0.20600318908691406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,32,balanced,0.08382399876912434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,1,power_law_1.2,0.21704959869384766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,32,balanced,0.08648000160853068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,1,power_law_1.2,0.22724480628967286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,32,balanced,0.1055413285891215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,1,power_law_1.2,0.24332160949707032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,32,balanced,0.13327466448148093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,32,balanced,0.19974400599797568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,1,power_law_1.2,0.28392319679260253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,32,balanced,0.2716533342997233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,1,power_law_1.2,0.31651840209960935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,32,balanced,0.2656799952189128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,1,power_law_1.2,0.39433600902557375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,32,balanced,0.2555306752522786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,1,power_law_1.2,0.47162880897521975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,32,balanced,0.2637653350830078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,1,power_law_1.2,0.6313471794128418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,32,balanced,0.2816320061683655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,32,balanced,0.25949867566426593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,1,power_law_1.2,0.7705088138580323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,32,balanced,0.2653813362121582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,1,power_law_1.2,1.0839103698730468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,32,balanced,0.295632004737854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,1,power_law_1.2,1.396224021911621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,32,balanced,0.06524266799290974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,32,balanced,0.27290666103363037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,1,power_law_1.2,2.016371154785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,32,balanced,0.06596800188223521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,32,balanced,0.06709866722424825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,32,balanced,0.27058666944503784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,1,power_law_1.2,2.6354496002197267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,32,balanced,0.07817066709200542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,32,balanced,0.27823466062545776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,32,balanced,0.09480533003807068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,1,power_law_1.2,3.2774784088134767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,32,balanced,0.13346133629480997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,32,balanced,0.2728586594263713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,1,power_law_1.2,5.174566268920898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,32,balanced,0.18352532386779785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,32,balanced,0.16893333196640015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,32,balanced,0.2914239962895711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,32,balanced,0.16901334126790366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,1,power_law_1.2,10.275968170166015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,32,balanced,0.16925867398579916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,32,balanced,0.16956265767415366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,32,balanced,0.29521600405375165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,32,balanced,0.1774133245150248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,32,balanced,0.17031999429066977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,32,balanced,0.18077866236368814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,32,balanced,0.18498667081197104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,32,balanced,0.17587200800577799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,32,balanced,0.19273066520690918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,32,balanced,0.19564799467722574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,32,balanced,0.206661323706309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,32,balanced,0.21665066480636597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,32,balanced,0.24758932987848917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,32,balanced,0.2828106681505839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,32,balanced,0.3241119980812073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,32,balanced,0.4110720157623291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,32,balanced,0.5620959997177124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,32,balanced,0.700767993927002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,32,balanced,0.9208587010701498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,32,balanced,0.06343466540177663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,32,balanced,0.06191466748714447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,32,balanced,0.06500266492366791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,32,balanced,0.07145600020885468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,32,balanced,0.09681066870689392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,32,balanced,0.13686399658521017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,32,balanced,0.16961065928141275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,32,balanced,0.15966932972272238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,32,balanced,0.17140799760818481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,32,balanced,0.16762133439381918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,32,balanced,0.17468265692392984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,32,balanced,0.16711999972661337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,32,balanced,0.15703999996185303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,32,balanced,0.17813867330551147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,32,balanced,0.1792959968249003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,32,balanced,0.1801813244819641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,32,balanced,1.178773323694865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,32,balanced,0.19001599152882895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,32,balanced,0.1824000080426534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,32,power_law_1.01,0.2044287919998169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,32,balanced,0.1945599913597107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,32,power_law_1.01,0.27102079391479494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,32,balanced,0.20268267393112183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,32,power_law_1.01,0.25914878845214845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,32,power_law_1.01,0.12451839447021484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,32,balanced,0.2205173373222351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,32,power_law_1.01,0.16522879600524903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,32,power_law_1.01,0.18176640272140504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,32,power_law_1.01,0.20076799392700195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,32,balanced,0.2418559988339742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,32,power_law_1.01,0.2151616096496582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,32,power_law_1.01,0.2466304063796997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,32,power_law_1.01,0.2294015884399414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,32,balanced,0.2700586716334025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,32,power_law_1.01,0.24063360691070557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,32,power_law_1.01,0.25070719718933104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,32,power_law_1.01,0.2575808048248291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,32,power_law_1.01,0.2585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,32,balanced,0.32493333021799725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,32,power_law_1.01,0.24997758865356445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,32,power_law_1.01,0.2411263942718506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,32,power_law_1.01,0.26494081020355226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,32,power_law_1.01,0.2597120046615601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,32,balanced,1.7887999216715496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,32,power_law_1.01,0.28097920417785643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,32,balanced,0.47039465109507245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,32,power_law_1.01,0.3155904054641724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,32,power_law_1.01,0.3613759994506836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,32,power_law_1.01,0.42818560600280764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,32,balanced,0.30290667215983075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,32,power_law_1.01,0.523737621307373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,32,power_law_1.01,0.6346303939819335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,32,balanced,0.3197653293609619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,32,power_law_1.01,0.8205887794494628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,32,balanced,0.5607200066248575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,32,power_law_1.01,1.0941823959350585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,32,balanced,0.3434986670811971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,32,power_law_1.01,1.286246395111084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,32,power_law_1.01,1.6752447128295898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,32,power_law_1.01,2.7120384216308593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,32,balanced,0.4041279951731364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,32,power_law_1.01,4.7757312774658205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,32,balanced,0.750661293665568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,32,balanced,0.5841973225275675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,32,balanced,0.6413120031356812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,32,balanced,0.9690079689025879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,32,balanced,0.8476373354593912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,32,balanced,3.3447093963623047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,32,balanced,1.136288007100423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,32,balanced,1.4517760276794434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,32,balanced,1.6821386019388835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,32,balanced,2.6733226776123047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,32,power_law_1.01,0.1288383960723877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,32,power_law_1.01,0.16219520568847656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,32,power_law_1.01,0.15507839918136596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,32,balanced,2.932058652242025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,32,power_law_1.01,0.09274880290031433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,32,power_law_1.01,0.10738559961318969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,32,power_law_1.01,0.12057600021362305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,32,power_law_1.01,0.12498559951782226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,32,power_law_1.01,0.15119359493255616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,32,power_law_1.01,0.14543360471725464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,32,power_law_1.01,0.15571199655532836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,32,power_law_1.01,0.14250880479812622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,32,power_law_1.01,0.15473920106887817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,32,power_law_1.01,0.1567296028137207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,32,power_law_1.01,0.15531519651412964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,32,power_law_1.01,0.16342400312423705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,32,power_law_1.01,0.1641152024269104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,32,power_law_1.01,0.18056960105895997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,32,power_law_1.01,0.18649599552154542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,32,power_law_1.01,0.20439679622650148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,32,power_law_1.01,0.22827520370483398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,32,power_law_1.01,0.2640320062637329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,32,power_law_1.01,0.34034559726715086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,32,power_law_1.01,0.4011839866638184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,32,power_law_1.01,0.13370879888534545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,32,power_law_1.01,0.5271935939788819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,32,power_law_1.01,0.1758911967277527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,32,power_law_1.01,0.6866432189941406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,32,power_law_1.01,0.1674496054649353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,32,power_law_1.01,0.9324607849121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,32,power_law_1.01,0.08892800211906433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,32,power_law_1.01,1.2012479782104493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,32,power_law_1.01,0.1044160008430481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,32,power_law_1.01,1.5256064414978028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,32,power_law_1.01,0.13022719621658324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,32,power_law_1.01,2.414944076538086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,32,power_law_1.01,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,32,power_law_1.01,0.15521279573440552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,32,power_law_1.01,4.3602752685546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,32,power_law_1.01,0.15240319967269897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,32,power_law_1.01,0.14776959419250488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,32,power_law_1.01,0.1477952003479004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,32,power_law_1.01,0.14897919893264772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,32,power_law_1.01,0.15704959630966187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,32,power_law_1.01,0.15902719497680665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,32,power_law_1.01,0.16036479473114013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,32,power_law_1.01,0.16455680131912231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,32,power_law_1.01,0.16586240530014038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,32,power_law_1.01,0.17418240308761596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,32,power_law_1.01,0.1890112042427063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,32,power_law_1.01,0.22456319332122804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,32,power_law_1.01,0.24445440769195556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,32,power_law_1.01,0.31007359027862547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,32,power_law_1.01,0.36538879871368407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,32,power_law_1.01,0.47983999252319337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,32,power_law_1.01,0.5837952136993408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,32,power_law_1.01,0.8051839828491211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,32,power_law_1.01,0.9955648422241211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,32,power_law_1.01,1.2600319862365723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,32,power_law_1.01,1.9829248428344726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,32,power_law_1.01,3.8600318908691404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,32,power_law_1.2,0.20860159397125244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,32,power_law_1.2,0.2592895984649658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,32,power_law_1.2,0.18727680444717407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,32,power_law_1.2,0.131660795211792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,32,power_law_1.2,0.14676480293273925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,32,power_law_1.2,0.17520639896392823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,32,power_law_1.2,0.19548159837722778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,32,power_law_1.2,0.22100479602813722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,32,power_law_1.2,0.21489920616149902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,32,power_law_1.2,0.1315392017364502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,32,power_law_1.2,0.2228991985321045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,32,power_law_1.2,0.16880639791488647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,32,power_law_1.2,0.24000000953674316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,32,power_law_1.2,0.16124800443649293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,32,power_law_1.2,0.23850879669189454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,32,power_law_1.2,0.08845440149307252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,32,power_law_1.2,0.2507839918136597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,32,power_law_1.2,0.10076160430908203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,32,power_law_1.2,0.24053120613098145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,32,power_law_1.2,0.12135039567947388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,32,power_law_1.2,0.2541887998580933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,32,power_law_1.2,0.12282880544662475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,32,power_law_1.2,0.22899839878082276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,32,power_law_1.2,0.14371199607849122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,32,power_law_1.2,0.24810879230499266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,32,power_law_1.2,0.15074559450149536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,32,power_law_1.2,0.2785599946975708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,32,power_law_1.2,0.14614399671554565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,32,power_law_1.2,0.30227839946746826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,32,power_law_1.2,0.14600319862365724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,32,power_law_1.2,0.3602560043334961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,32,power_law_1.2,0.15352959632873536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,32,power_law_1.2,0.37817599773406985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,32,power_law_1.2,0.1559615969657898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,32,power_law_1.2,0.5066880226135254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,32,power_law_1.2,0.15410560369491577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,32,power_law_1.2,0.17330559492111205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,32,power_law_1.2,0.5536896228790283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,32,power_law_1.2,0.7037439823150635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,32,power_law_1.2,0.1764415979385376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,32,power_law_1.2,0.1709439992904663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,32,power_law_1.2,0.9675775527954101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,32,power_law_1.2,0.1926911950111389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,32,power_law_1.2,1.317356777191162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,32,power_law_1.2,0.20563840866088867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,32,power_law_1.2,1.6915071487426758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,32,power_law_1.2,0.25396480560302737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,32,power_law_1.2,2.002163124084473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,32,power_law_1.2,0.2979007959365845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,32,power_law_1.2,3.342438507080078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,32,power_law_1.2,0.38146560192108153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,32,power_law_1.2,0.43912320137023925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,32,power_law_1.2,6.3088127136230465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,32,power_law_1.2,0.5992000102996826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,32,power_law_1.2,0.738431978225708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,32,power_law_1.2,1.1286144256591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,32,power_law_1.2,0.1462399959564209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,32,power_law_1.2,1.335641574859619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,32,power_law_1.2,0.18380160331726075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,32,power_law_1.2,0.16115200519561768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,32,power_law_1.2,1.6592704772949218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,32,power_law_1.2,0.09410560131072998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,32,power_law_1.2,2.4733375549316405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,32,power_law_1.2,0.10215679407119752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,32,power_law_1.2,5.510649490356445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,32,power_law_1.2,0.12165759801864624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,32,power_law_1.2,0.12237440347671509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,32,power_law_1.2,0.14769920110702514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,32,power_law_1.2,0.1437824010848999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,32,power_law_1.2,0.15966720581054689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,32,power_law_1.2,0.15003520250320435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,32,power_law_1.2,0.1590783953666687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,32,power_law_1.2,0.15765759944915772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,32,power_law_1.2,0.14866559505462645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,32,power_law_1.2,0.15957119464874267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,32,power_law_1.2,0.17051520347595214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,32,power_law_1.2,0.16419199705123902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,32,power_law_1.2,0.19639040231704713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,32,power_law_1.2,0.2076159954071045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,32,power_law_1.2,0.24335360527038574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,32,power_law_1.2,0.2751104116439819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,32,power_law_1.2,0.3356031894683838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,32,power_law_1.2,0.3897408008575439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,32,power_law_1.2,0.5272768020629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,32,power_law_1.2,0.6428415775299072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,32,power_law_1.2,0.8618304252624511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,16,balanced,0.06738133231798808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,16,balanced,0.07126399874687195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,32,power_law_1.2,1.2296319961547852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,16,balanced,0.07303999861081441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,16,balanced,0.09749333063761394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,32,power_law_1.2,1.468832015991211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,16,balanced,0.12586667140324911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,16,balanced,0.19476266702016196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,32,power_law_1.2,2.134783935546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,16,balanced,0.24939733743667603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,16,balanced,0.25338133176167804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,16,balanced,0.24065067370732626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,32,power_law_1.2,4.936902236938477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,16,balanced,0.2500159939130147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,16,balanced,0.2601173321406047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,16,balanced,0.2569653391838074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,16,balanced,0.24270933866500854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,16,balanced,0.24837867418924967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,16,balanced,0.25602134068806964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,16,balanced,0.24835733572642008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,16,balanced,0.26985599597295123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,16,balanced,0.26045332352320355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,16,balanced,0.27669866879781085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,16,balanced,0.28008532524108887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,16,balanced,0.30290667215983075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,16,balanced,0.3255680004755656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,16,balanced,0.35336001714070636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,16,balanced,0.41655464967091876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,16,balanced,0.5994879802068075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,16,balanced,0.6706879933675131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,16,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,16,balanced,0.055573334296544395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,16,balanced,0.06163733204205831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,16,balanced,0.0758186678091685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,16,balanced,0.08776533603668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,16,balanced,0.13317333658536276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,16,balanced,0.1702079971631368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,16,balanced,0.1650879979133606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,16,balanced,0.1642453372478485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,16,balanced,0.887343962987264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,16,balanced,0.16185067097345987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,16,balanced,0.155349334081014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,16,balanced,0.16290133198102316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,16,balanced,0.16269333163897196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,16,balanced,0.16532267133394876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,16,balanced,0.057205334305763245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,16,balanced,0.058778668443361916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,16,balanced,0.17683200041453043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,16,balanced,0.06533866624037425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,16,balanced,0.17004799842834473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,16,balanced,0.0735999991496404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,16,balanced,0.10413333773612976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,16,balanced,0.18049599726994833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,16,balanced,0.13428800304730734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,16,balanced,0.16365333398183188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,16,balanced,0.189626673857371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,16,balanced,0.17200533548990884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,16,balanced,0.20266133546829224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,16,balanced,0.1691733400026957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,16,balanced,0.16514133413632712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,16,balanced,1.173413356145223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,16,balanced,0.15970133741696677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,16,balanced,0.22394132614135742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,16,balanced,0.16460800170898438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,16,balanced,0.16139733791351318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,16,balanced,0.2434026598930359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,16,balanced,0.16358400384585062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,16,balanced,0.17627199490865073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,16,balanced,0.1643786629041036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,16,balanced,0.288975993792216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,16,balanced,0.17957866191864014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,16,balanced,0.1808533271153768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,16,balanced,0.19110933939615884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,16,balanced,0.33448533217112225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,16,balanced,0.20572266976038614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,16,balanced,0.22023467222849527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,16,balanced,0.42739200592041016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,16,balanced,0.25144533316294354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,16,balanced,1.7472160657246907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,16,balanced,0.28015466531117755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,16,balanced,0.5727306604385376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,16,balanced,0.35045333703358966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,16,power_law_1.01,0.13025280237197875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,16,power_law_1.01,0.20327680110931395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,16,power_law_1.01,0.24428160190582277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,16,power_law_1.01,0.10702719688415527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,16,power_law_1.01,0.12892160415649415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,16,balanced,0.49032533168792725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,16,power_law_1.01,0.1793023943901062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,16,power_law_1.01,0.17086080312728882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,16,balanced,0.7422560056050619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,16,power_law_1.01,0.21600639820098877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,16,power_law_1.01,0.20214400291442872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,16,power_law_1.01,0.21634559631347655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,16,power_law_1.01,0.23036160469055175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,16,balanced,0.6115786631902059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,16,power_law_1.01,0.2208575963973999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,16,power_law_1.01,0.2080832004547119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,16,power_law_1.01,0.24122240543365478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,16,power_law_1.01,0.2354048013687134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,16,power_law_1.01,0.23705599308013917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,16,power_law_1.01,0.24662399291992188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,16,balanced,0.9670560359954834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,16,power_law_1.01,0.25622398853302003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,16,power_law_1.01,0.26645119190216066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,16,power_law_1.01,0.292960000038147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,16,balanced,0.8049120108286539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,16,power_law_1.01,0.3315648078918457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,16,power_law_1.01,0.406771183013916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,16,power_law_1.01,0.45493121147155763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,16,power_law_1.01,0.5893824100494385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,16,power_law_1.01,0.70447998046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,16,power_law_1.01,1.0215999603271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,16,balanced,3.1094239552815757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,16,power_law_1.01,1.2406847953796387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,16,power_law_1.01,1.4199359893798829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,16,balanced,1.0247039794921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,16,power_law_1.01,2.285004806518555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,16,balanced,1.223861296971639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,16,power_law_1.01,4.3102272033691404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,16,balanced,1.5559360186258953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,16,balanced,1.8673653602600098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,16,balanced,2.9102560679117837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,16,balanced,3.5551681518554688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,16,power_law_1.01,0.09253759980201721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,16,power_law_1.01,0.1291264057159424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,16,power_law_1.01,0.17048319578170776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,16,power_law_1.01,0.10652159452438355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,16,power_law_1.01,0.07883520126342773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,16,power_law_1.01,0.1444607973098755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,16,power_law_1.01,0.09621120095252991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,16,power_law_1.01,0.1763200044631958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,16,power_law_1.01,0.11775360107421876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,16,power_law_1.01,0.07983360290527344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,16,power_law_1.01,0.10359040498733521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,16,power_law_1.01,0.1251711964607239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,16,power_law_1.01,0.11249920129776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,16,power_law_1.01,0.1351680040359497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,16,power_law_1.01,0.1252351999282837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,16,power_law_1.01,0.13920639753341674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,16,power_law_1.01,0.14647040367126465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,16,power_law_1.01,0.1451200008392334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,16,power_law_1.01,0.1474496006965637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,16,power_law_1.01,0.13973759412765502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,16,power_law_1.01,0.1493183970451355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,16,power_law_1.01,0.14985599517822265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,16,power_law_1.01,0.1482624053955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,16,power_law_1.01,0.14456959962844848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,16,power_law_1.01,0.1553663969039917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,16,power_law_1.01,0.15941760540008545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,16,power_law_1.01,0.1522047996520996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,16,power_law_1.01,0.1540735960006714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,16,power_law_1.01,0.15380480289459228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,16,power_law_1.01,0.1562432050704956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,16,power_law_1.01,0.16669440269470215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,16,power_law_1.01,0.17207679748535157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,16,power_law_1.01,0.15962239503860473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,16,power_law_1.01,0.17902079820632935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,16,power_law_1.01,0.1677440047264099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,16,power_law_1.01,0.20066559314727783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,16,power_law_1.01,0.17428480386734008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,16,power_law_1.01,0.22965760231018068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,16,power_law_1.01,0.19139200448989868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,16,power_law_1.01,0.27044479846954345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,16,power_law_1.01,0.21735680103302002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,16,power_law_1.01,0.326310396194458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,16,power_law_1.01,0.24252159595489503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,16,power_law_1.01,0.39426560401916505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,16,power_law_1.01,0.3018111944198608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,16,power_law_1.01,0.5237887859344482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,16,power_law_1.01,0.3537215948104858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,16,power_law_1.01,0.6603968143463135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,16,power_law_1.01,0.46581759452819826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,16,power_law_1.01,0.9286591529846191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,16,power_law_1.01,0.5625088214874268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,16,power_law_1.01,1.21081600189209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,16,power_law_1.01,0.8065600395202637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,16,power_law_1.01,1.4460160255432128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,16,power_law_1.01,1.043008041381836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,16,power_law_1.01,2.311545562744141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,16,power_law_1.01,1.2585599899291993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,16,power_law_1.01,4.532889556884766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,16,power_law_1.01,1.9229951858520509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,16,power_law_1.01,3.634675216674805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,16,power_law_1.2,0.1252671957015991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,16,power_law_1.2,0.1904255986213684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,16,power_law_1.2,0.1771775960922241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,16,power_law_1.2,0.10314240455627441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,16,power_law_1.2,0.13586560487747193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,16,power_law_1.2,0.17122559547424315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,16,power_law_1.2,0.1684928059577942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,16,power_law_1.2,0.1925503969192505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,16,power_law_1.2,0.19029120206832886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,16,power_law_1.2,0.19508479833602904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,16,power_law_1.2,0.20590078830718994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,16,power_law_1.2,0.21616001129150392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,16,power_law_1.2,0.22085120677947997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,16,power_law_1.2,0.21831679344177246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,16,power_law_1.2,0.08903679847717286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,16,power_law_1.2,0.23171839714050294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,16,power_law_1.2,0.12581759691238403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,16,power_law_1.2,0.22062718868255615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,16,power_law_1.2,0.23284480571746827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,16,power_law_1.2,0.10807679891586304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,16,power_law_1.2,0.2644351959228516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,16,power_law_1.2,0.07549440264701843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,16,power_law_1.2,0.2755520105361938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,16,power_law_1.2,0.08717439770698547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,16,power_law_1.2,0.3313920021057129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,16,power_law_1.2,0.10690560340881347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,16,power_law_1.2,0.3555903911590576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,16,power_law_1.2,0.11707520484924316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,16,power_law_1.2,0.41947522163391116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,16,power_law_1.2,0.1321023941040039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,16,power_law_1.2,0.47416319847106936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,16,power_law_1.2,0.13569920063018798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,16,power_law_1.2,0.7034048080444336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,16,power_law_1.2,0.1379520058631897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,16,power_law_1.2,0.7444096088409424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,16,power_law_1.2,0.1461632013320923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,16,power_law_1.2,1.0085887908935547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,16,power_law_1.2,0.14294400215148925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,16,power_law_1.2,0.14884480237960815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,16,power_law_1.2,1.3226816177368164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,16,power_law_1.2,0.1466879963874817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,16,power_law_1.2,1.6907712936401367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,16,power_law_1.2,0.16467200517654418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,16,power_law_1.2,2.612435150146484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,16,power_law_1.2,0.16603519916534423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,16,power_law_1.2,5.403942489624024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,16,power_law_1.2,0.17172479629516602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,16,power_law_1.2,0.18453119993209838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,16,power_law_1.2,0.20437119007110596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,16,power_law_1.2,0.24590721130371093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,16,power_law_1.2,0.2828991889953613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,16,power_law_1.2,0.34461441040039065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,16,power_law_1.2,0.4289535999298096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,16,power_law_1.2,0.5505407810211181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,16,power_law_1.2,0.6890687942504883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,16,power_law_1.2,1.055679988861084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,16,power_law_1.2,1.3506752014160157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,16,power_law_1.2,1.6059904098510742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,16,power_law_1.2,2.4139263153076174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,16,power_law_1.2,4.841228866577149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,16,power_law_1.2,0.10214400291442871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,16,power_law_1.2,0.13815040588378907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,8,balanced,0.06122133135795593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,8,balanced,0.0655626654624939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,16,power_law_1.2,0.12295039892196655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,8,balanced,0.0799786647160848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,8,balanced,0.09409067034721375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,16,power_law_1.2,0.0834879994392395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,8,balanced,0.12009066343307495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,16,power_law_1.2,0.09452800154685974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,8,balanced,0.1946773330370585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,8,balanced,0.25861332813898724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,16,power_law_1.2,0.10447360277175903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,8,balanced,0.2355840007464091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,8,balanced,0.2427306572596232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,16,power_law_1.2,0.12583680152893068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,8,balanced,0.24252265691757202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,16,power_law_1.2,0.14030079841613768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,8,balanced,0.25328532854715985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,8,balanced,0.2552479902903239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,16,power_law_1.2,0.1473536014556885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,8,balanced,0.24103466669718424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,16,power_law_1.2,0.14017280340194702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,8,balanced,0.23632532358169556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,16,power_law_1.2,0.14936959743499756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,8,balanced,0.2511146664619446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,16,power_law_1.2,0.15009280443191528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,8,balanced,0.25749866167704266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,8,balanced,0.25805334250132245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,16,power_law_1.2,0.15932799577713014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,8,balanced,0.2670666575431824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,16,power_law_1.2,0.15710079669952393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,16,power_law_1.2,0.16286079883575438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,8,balanced,0.27007466554641724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,16,power_law_1.2,0.16153600215911865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,8,balanced,0.2892640034357707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,16,power_law_1.2,0.16678400039672853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,16,power_law_1.2,0.18791680335998534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,8,balanced,0.31140265862147015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,16,power_law_1.2,0.19714560508728027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,16,power_law_1.2,0.23809919357299805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,8,balanced,0.34223465124766034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,16,power_law_1.2,0.2620800018310547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,16,power_law_1.2,0.3385024070739746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,16,power_law_1.2,0.3934528112411499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,8,balanced,0.37998934586842853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,16,power_law_1.2,0.5078144073486328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,16,power_law_1.2,0.6272448062896728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,16,power_law_1.2,0.895263957977295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,16,power_law_1.2,1.0922368049621582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,8,balanced,0.45864001909891766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,16,power_law_1.2,1.4960960388183593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,16,power_law_1.2,2.190700721740723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,16,power_law_1.2,4.38263053894043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,8,balanced,0.6249333222707113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,8,balanced,0.7583786646525065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,8,balanced,0.05217066903909048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,8,balanced,0.05533866584300995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,8,balanced,0.05938666562239329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,8,balanced,0.07407466570536296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,8,balanced,0.08881066242853801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,8,balanced,0.13102933764457703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,8,balanced,0.16456000010172525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,8,balanced,0.168229341506958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,8,balanced,0.15745066603024802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,8,balanced,0.15606932838757834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,8,balanced,0.15877333283424377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,8,balanced,0.16057067116101584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,8,balanced,0.9806400140126547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,8,balanced,0.16179733475049338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,8,balanced,0.1618613302707672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,8,balanced,0.17242666085561117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,8,balanced,0.1741973360379537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,8,balanced,0.17994133631388345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,8,balanced,0.1937546730041504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,8,balanced,0.21049066384633383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,8,balanced,0.2365493377049764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,8,balanced,0.2592746615409851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,8,balanced,0.05449600021044413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,8,balanced,0.06238933404286703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,8,balanced,0.0682826687892278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,8,balanced,1.2680266698201497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,8,balanced,0.31639466683069867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,8,balanced,0.07844266792138417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,8,balanced,0.10786133011182149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,8,balanced,0.15436800320943198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,8,balanced,0.1914986570676168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,8,balanced,0.3661866585413615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,8,balanced,0.19298134247461954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,8,balanced,0.19020267327626547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,8,balanced,0.1860533356666565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,8,balanced,0.179530660311381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,8,balanced,0.17308799425760904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,8,balanced,0.47833065191904706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,8,balanced,0.17667200167973837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,8,balanced,0.1813653310139974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,8,balanced,0.18625599145889282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,8,balanced,0.1824480096499125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,8,balanced,0.1888479987780253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,8,balanced,0.6302933295567831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,8,balanced,0.19818667570749918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,8,balanced,0.20637865861256918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,8,balanced,0.22579733530680338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,8,balanced,0.24460800488789877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,8,balanced,0.8394827047983805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,8,balanced,0.2848479946454366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,8,balanced,1.8872106870015461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,8,balanced,0.3259999950726827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,8,balanced,1.0960586865743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,8,balanced,0.41154666741689044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,8,balanced,0.5624639987945557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,8,balanced,1.355610688527425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,8,balanced,0.727509339650472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,8,power_law_1.01,0.09488000273704529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,8,power_law_1.01,0.12284799814224243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,8,power_law_1.01,0.1964416027069092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,8,power_law_1.01,0.09535359740257263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,8,balanced,0.9588053226470947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,8,power_law_1.01,0.11564160585403442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,8,power_law_1.01,0.142848002910614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,8,balanced,2.0868746439615884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,8,power_law_1.01,0.16257280111312866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,8,power_law_1.01,0.18491519689559938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,8,power_law_1.01,0.18511359691619872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,8,power_law_1.01,0.19266560077667236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,8,balanced,3.4908053080240884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,8,power_law_1.01,0.19864319562911986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,8,power_law_1.01,0.20371840000152588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,8,power_law_1.01,0.2100543975830078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,8,power_law_1.01,0.2217087984085083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,8,balanced,1.2027733325958252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,8,power_law_1.01,0.21957120895385743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,8,power_law_1.01,0.24018559455871583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,8,power_law_1.01,0.23995521068572997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,8,power_law_1.01,0.2530431985855103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,8,power_law_1.01,0.27078399658203123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,8,power_law_1.01,0.2978624105453491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,8,power_law_1.01,0.3225919961929321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,8,power_law_1.01,0.3943295955657959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,8,power_law_1.01,0.44899840354919435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,8,power_law_1.01,0.6071807861328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,8,power_law_1.01,0.7441535949707031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,8,power_law_1.01,0.9977151870727539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,8,power_law_1.01,1.176313591003418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,8,power_law_1.01,1.4637056350708009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,8,balanced,1.8285226821899414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,8,power_law_1.01,2.2271551132202148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,8,power_law_1.01,4.32716178894043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,8,balanced,4.000634511311849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,8,balanced,3.483994801839193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,8,power_law_1.01,0.073990398645401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,8,power_law_1.01,0.09327359795570374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,8,power_law_1.01,0.12907520532608033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,8,power_law_1.01,0.075135999917984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,8,power_law_1.01,0.08714240193367004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,8,power_law_1.01,0.11198079586029053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,8,power_law_1.01,0.11634559631347656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,8,power_law_1.01,0.12856320142745972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,8,power_law_1.01,0.13514879941940308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,8,power_law_1.01,0.1420032024383545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,8,power_law_1.01,0.14122240543365477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,8,power_law_1.01,0.14732160568237304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,8,power_law_1.01,0.1487104058265686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,8,power_law_1.01,0.1495743989944458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,8,power_law_1.01,0.16058239936828614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,8,power_law_1.01,0.16976640224456788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,8,power_law_1.01,0.08300160169601441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,8,power_law_1.01,0.17805440425872804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,8,power_law_1.01,0.10855679512023926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,8,power_law_1.01,0.18999680280685424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,8,power_law_1.2,0.09467520117759705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,8,power_law_1.01,0.15096960067749024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,8,power_law_1.01,0.20959360599517823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,8,power_law_1.2,0.12421120405197143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,8,power_law_1.01,0.08459519743919372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,8,power_law_1.01,0.24134399890899658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,8,power_law_1.2,0.11228159666061402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,8,power_law_1.01,0.10144000053405762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,8,power_law_1.01,0.2718400001525879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,8,power_law_1.2,0.09831680059432983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,8,power_law_1.01,0.11931519508361817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,8,power_law_1.01,0.33909120559692385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,8,power_law_1.2,0.10964479446411132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,8,power_law_1.01,0.13189760446548462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,8,power_law_1.01,0.41150717735290526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,8,power_law_1.2,0.12780799865722656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,8,power_law_1.01,0.14555519819259644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,8,power_law_1.2,0.1523327946662903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,8,power_law_1.01,0.1537727952003479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,8,power_law_1.01,0.552620792388916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,8,power_law_1.01,0.15117440223693848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,8,power_law_1.2,0.17359999418258668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,8,power_law_1.01,0.6911359786987304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,8,power_law_1.01,0.15161600112915039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,8,power_law_1.2,0.18131840229034424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,8,power_law_1.01,0.9584768295288086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,8,power_law_1.01,0.15983999967575074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,8,power_law_1.2,0.18273279666900635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,8,power_law_1.01,1.2988672256469727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,8,power_law_1.01,0.163756799697876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,8,power_law_1.2,0.20030720233917237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,8,power_law_1.01,1.5867903709411622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,8,power_law_1.01,0.16368000507354735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,8,power_law_1.2,0.20162560939788818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,8,power_law_1.01,0.1662335991859436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,8,power_law_1.01,2.4147775650024412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,8,power_law_1.2,0.1980288028717041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,8,power_law_1.01,0.1719231963157654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,8,power_law_1.2,0.2166719913482666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,8,power_law_1.01,4.674764633178711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,8,power_law_1.01,0.1823040008544922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,8,power_law_1.2,0.22722558975219725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,8,power_law_1.01,0.19545600414276124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,8,power_law_1.2,0.22617599964141846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,8,power_law_1.01,0.20507519245147704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,8,power_law_1.2,0.24109439849853515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,8,power_law_1.01,0.23744640350341797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,8,power_law_1.2,0.2611840009689331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,8,power_law_1.01,0.2699199914932251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,8,power_law_1.2,0.27384960651397705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,8,power_law_1.01,0.34034559726715086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,8,power_law_1.2,0.32457599639892576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,8,power_law_1.01,0.3914880037307739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,8,power_law_1.2,0.349401593208313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,8,power_law_1.01,0.5174528121948242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,8,power_law_1.2,0.41144962310791017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,8,power_law_1.01,0.6301631927490234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,8,power_law_1.2,0.4774655818939209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,8,power_law_1.01,0.8838656425476075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,8,power_law_1.2,0.6266240119934082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,8,power_law_1.01,1.1285632133483887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,8,power_law_1.2,0.7227456092834472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,8,power_law_1.01,1.3822976112365724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,8,power_law_1.2,1.0109312057495117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,8,power_law_1.01,2.0531391143798827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,8,power_law_1.2,1.2786944389343262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,8,power_law_1.2,0.07331839799880982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,8,power_law_1.2,1.5704192161560058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,8,power_law_1.2,0.09331200122833253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,8,power_law_1.01,4.325088119506836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,8,power_law_1.2,0.10154880285263061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,8,power_law_1.2,2.597599983215332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,8,power_law_1.2,0.0752128005027771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,8,power_law_1.2,5.130956649780273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,8,power_law_1.2,0.08776320219039917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,8,power_law_1.2,0.09687039852142335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,8,power_law_1.2,0.11345920562744141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,8,power_law_1.2,0.12439039945602418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,8,power_law_1.2,0.12395520210266113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,8,power_law_1.2,0.1371392011642456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,8,power_law_1.2,0.13983999490737914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,8,power_law_1.2,0.14991359710693358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,8,power_law_1.2,0.1450495958328247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,8,power_law_1.2,0.14725760221481324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,8,power_law_1.2,0.16179200410842895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,8,power_law_1.2,0.16409599781036377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,8,power_law_1.2,0.17777279615402222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,8,power_law_1.2,0.19429759979248046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,8,power_law_1.2,0.21250560283660888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,8,power_law_1.2,0.24438400268554689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,8,power_law_1.2,0.29125759601593015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,8,power_law_1.2,0.35320320129394533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,8,power_law_1.2,0.4297791957855225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,8,power_law_1.2,0.5816127777099609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,8,power_law_1.2,0.7277056217193604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,8,power_law_1.2,1.0129280090332031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,8,power_law_1.2,1.3609279632568358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,8,power_law_1.2,1.6254335403442384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,8,power_law_1.2,2.796499252319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,8,power_law_1.2,5.005382537841797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,8,power_law_1.2,0.08465920090675354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,8,power_law_1.2,0.10675200223922729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,8,power_law_1.2,0.10776959657669068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,8,power_law_1.2,0.08503680229187012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,8,power_law_1.2,0.09655680060386658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,8,power_law_1.2,0.11271040439605713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,8,power_law_1.2,0.12187520265579224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,8,power_law_1.2,0.14213759899139405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,8,power_law_1.2,0.1484671950340271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,4,balanced,0.0625493327776591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,8,power_law_1.2,0.14476799964904785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,4,balanced,0.06693333387374878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,4,balanced,0.07916800181070964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,8,power_law_1.2,0.15751680135726928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,4,balanced,0.10077866911888123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,4,balanced,0.12808533509572348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,8,power_law_1.2,0.16198400259017945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,4,balanced,0.20085867245992026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,4,balanced,0.26198933521906537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,8,power_law_1.2,0.15635839700698853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,4,balanced,0.05454933146635691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,4,balanced,0.2613919973373413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,4,balanced,0.05865600208441416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,8,power_law_1.2,0.16046080589294434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,4,balanced,0.06314133107662201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,4,balanced,0.2632426619529724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,4,balanced,0.07786133388678233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,4,balanced,0.2541813254356384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,8,power_law_1.2,0.17422080039978027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,4,balanced,0.10518933335940044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,4,balanced,0.259168008963267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,4,balanced,0.15125333269437155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,8,power_law_1.2,0.17946879863739013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,4,balanced,0.26049600044886273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,4,balanced,0.18890666961669922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,4,balanced,0.2507680058479309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,8,power_law_1.2,0.19285119771957399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,4,balanced,0.18945600589116415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,4,balanced,0.25759466489156085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,4,balanced,0.19011199474334717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,8,power_law_1.2,0.19520000219345093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,4,balanced,0.26521599292755127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,4,balanced,0.18184000253677368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,8,power_law_1.2,0.22218880653381348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,4,balanced,0.2605920036633809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,4,balanced,0.18624534209569296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,4,balanced,0.1874720056851705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,4,balanced,0.2664053241411845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,8,power_law_1.2,0.24134399890899658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,4,balanced,0.18620266517003378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,8,power_law_1.2,0.27546238899230957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,4,balanced,0.28085867563883465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,4,balanced,0.19269333283106485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,4,balanced,0.1991306742032369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,4,balanced,0.2948746681213379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,8,power_law_1.2,0.34760959148406984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,4,balanced,0.20005333423614502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,8,power_law_1.2,0.4077439785003662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,4,balanced,0.3205866614977519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,4,balanced,0.21166932582855225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,8,power_law_1.2,0.5582464218139649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,4,balanced,0.22714134057362875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,4,balanced,0.3444053332010905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,8,power_law_1.2,0.6934463977813721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,4,balanced,0.24288000663121542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,8,power_law_1.2,0.9327936172485352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,4,balanced,0.3967306613922119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,4,balanced,0.2784000039100647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,8,power_law_1.2,1.1861568450927735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,8,power_law_1.2,1.610348892211914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,4,balanced,0.3110559980074565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,4,balanced,0.4496106704076131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,8,power_law_1.2,2.2489856719970702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,4,balanced,0.37858132521311444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,8,power_law_1.2,4.810176086425781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,4,balanced,0.5574880043665568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,4,balanced,0.4456746578216553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,4,balanced,0.5847733418146769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,4,balanced,0.7564693291982015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,4,balanced,0.7791840235392252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,4,balanced,0.060831998785336815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,4,balanced,0.06771199901898702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,4,balanced,0.9663573106129965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,4,balanced,0.07394133508205414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,4,balanced,0.08932266632715861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,4,balanced,0.12315199772516887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,4,balanced,0.18247467279434204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,4,balanced,0.22961600621541342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,4,balanced,0.23404266436894736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,4,balanced,1.0514559745788574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,4,balanced,0.22103466590245566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,4,balanced,0.22167466084162393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,4,balanced,0.21964800357818604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,4,balanced,0.21837333838144937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,4,balanced,0.220960001150767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,4,balanced,0.22521066665649414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,4,balanced,1.2450079917907715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,4,balanced,0.2319200038909912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,4,balanced,0.22946133216222128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,4,balanced,0.23873066902160645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,4,balanced,1.3701386451721191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,4,balanced,0.2542293270428975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,4,balanced,0.26436267296473187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,4,balanced,0.2919519941012065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,4,balanced,0.31652265787124634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,4,balanced,1.5504053433736165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,4,balanced,0.3727253278096517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,4,balanced,1.6943519910176594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,4,balanced,0.4256266752878825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,4,balanced,0.5469706853230795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,4,balanced,0.7506133715311686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,4,balanced,2.3341546058654785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,4,balanced,2.6010079383850098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,4,balanced,0.9699359734853109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,4,balanced,1.286197344462077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,4,balanced,1.606160004933675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,4,balanced,4.3818667729695635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,4,balanced,5.012415885925293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,4,power_law_1.01,0.0812928020954132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,4,power_law_1.01,0.1007040023803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,4,power_law_1.01,0.13082239627838135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,4,power_law_1.01,0.09281280040740966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,4,power_law_1.01,0.1155776023864746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,4,power_law_1.01,0.14149119853973388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,4,power_law_1.01,0.16620800495147706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,4,power_law_1.01,0.1958847999572754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,4,balanced,2.4581119219462075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,4,power_law_1.01,0.19794559478759766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,4,power_law_1.01,0.19733760356903077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,4,power_law_1.01,0.20261120796203613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,4,power_law_1.01,0.2093440055847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,4,power_law_1.01,0.21921920776367188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,4,power_law_1.01,0.22369918823242188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,4,power_law_1.01,0.2336575984954834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,4,power_law_1.01,0.23708798885345458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,4,power_law_1.01,0.25591681003570554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,4,power_law_1.01,0.2744960069656372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,4,power_law_1.01,0.29307520389556885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,4,power_law_1.01,0.3243839979171753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,4,power_law_1.01,0.3524032115936279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,4,power_law_1.01,0.4345088005065918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,4,power_law_1.01,0.4988800048828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,4,power_law_1.01,0.6445119857788086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,4,power_law_1.01,0.7926335811614991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,4,power_law_1.01,1.102182388305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,4,power_law_1.01,1.4025152206420899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,4,power_law_1.01,1.7562559127807618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,4,power_law_1.01,2.557062339782715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,4,power_law_1.01,4.9812671661376955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,4,balanced,4.697093327840169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,4,power_law_1.01,0.06757760047912598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,4,power_law_1.01,0.08171520233154297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,4,power_law_1.01,0.07587839961051941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,4,power_law_1.01,0.10637439489364624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,4,power_law_1.01,0.09119359850883484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,4,power_law_1.01,0.08074880242347718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,4,power_law_1.01,0.12248320579528808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,4,power_law_1.01,0.09493119716644287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,4,power_law_1.01,0.0880895972251892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,4,power_law_1.01,0.12071679830551148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,4,power_law_1.01,0.10579839944839478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,4,power_law_1.01,0.13055360317230225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,4,power_law_1.01,0.1389631986618042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,4,power_law_1.01,0.1484992027282715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,4,power_law_1.01,0.14449280500411987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,4,power_law_1.01,0.14841599464416505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,4,power_law_1.01,0.16864000558853148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,4,power_law_1.01,0.1549888014793396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,4,power_law_1.01,0.1712767958641052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,4,power_law_1.01,0.15374079942703248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,4,power_law_1.01,0.15839999914169312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,4,power_law_1.01,0.17305599451065062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,4,power_law_1.01,0.16229759454727172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,4,power_law_1.01,0.184716796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,4,power_law_1.01,0.17612160444259645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,4,power_law_1.01,0.18494080305099486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,4,power_law_1.01,0.18213119506835937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,4,power_law_1.01,0.193503999710083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,4,power_law_1.01,0.18202240467071534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,4,power_law_1.01,0.1938815951347351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,4,power_law_1.01,0.20280959606170654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,4,power_law_1.01,0.20395519733428955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,4,power_law_1.01,0.21765120029449464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,4,power_law_1.01,0.21422719955444336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,4,power_law_1.01,0.23800959587097167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,4,power_law_1.01,0.21864960193634034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,4,power_law_1.01,0.2821120023727417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,4,power_law_1.01,0.23917438983917236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,4,power_law_1.01,0.31409919261932373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,4,power_law_1.01,0.24792959690093994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,4,power_law_1.01,0.4108799934387207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,4,power_law_1.01,0.2871999979019165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,4,power_law_1.01,0.49283199310302733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,4,power_law_1.01,0.3331264019012451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,4,power_law_1.01,0.4139711856842041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,4,power_law_1.01,0.6505856037139892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,4,power_law_1.01,0.488921594619751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,4,power_law_1.01,0.8112575531005859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,4,power_law_1.01,0.6538496017456055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,4,power_law_1.01,1.1794624328613281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,4,power_law_1.01,0.8084544181823731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,4,power_law_1.01,1.4849472045898438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,4,power_law_1.01,1.1038144111633301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,4,power_law_1.01,1.8256128311157227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,4,power_law_1.01,1.4385791778564454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,4,power_law_1.01,2.787507247924805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,4,power_law_1.01,1.7462528228759766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,4,power_law_1.01,5.467782211303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,4,power_law_1.01,2.654694366455078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,4,power_law_1.01,5.267961502075195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,4,power_law_1.2,0.08469120264053345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,4,power_law_1.2,0.10289920568466186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,4,power_law_1.2,0.106387197971344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,4,power_law_1.2,0.0946943998336792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,4,power_law_1.2,0.11257599592208863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,4,power_law_1.2,0.1341055989265442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,4,power_law_1.2,0.16230399608612062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,4,power_law_1.2,0.17947520017623902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,4,power_law_1.2,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,4,power_law_1.2,0.19281920194625854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,4,power_law_1.2,0.08102399706840516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,4,power_law_1.2,0.19162240028381347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,4,power_law_1.2,0.07504000067710877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,4,power_law_1.2,0.20619521141052247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,4,power_law_1.2,0.08010879755020142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,4,power_law_1.2,0.20891520977020264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,4,power_law_1.2,0.09108480215072631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,4,power_law_1.2,0.21162879467010498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,4,power_law_1.2,0.11089919805526734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,4,power_law_1.2,0.21818239688873292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,4,power_law_1.2,0.22819840908050537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,4,power_law_1.2,0.12017920017242431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,4,power_law_1.2,0.2525696039199829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,4,power_law_1.2,0.14072959423065184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,4,power_law_1.2,0.2614527940750122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,4,power_law_1.2,0.15017600059509278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,4,power_law_1.2,0.28327040672302245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,4,power_law_1.2,0.1475648045539856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,4,power_law_1.2,0.2983936071395874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,4,power_law_1.2,0.14945919513702394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,4,power_law_1.2,0.3364288091659546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,4,power_law_1.2,0.15968639850616456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,4,power_law_1.2,0.36455678939819336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,4,power_law_1.2,0.16320639848709106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,4,power_law_1.2,0.4549248218536377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,4,power_law_1.2,0.16547199487686157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,4,power_law_1.2,0.5222080230712891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,4,power_law_1.2,0.18413439989089966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,4,power_law_1.2,0.6858816146850586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,4,power_law_1.2,0.1860479950904846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,4,power_law_1.2,0.19535360336303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,4,power_law_1.2,0.803660774230957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,4,power_law_1.2,0.22322559356689453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,4,power_law_1.2,1.1439680099487304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,4,power_law_1.2,0.24543359279632568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,4,power_law_1.2,1.484716796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,4,power_law_1.2,0.28153600692749026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,4,power_law_1.2,1.7836927413940429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,4,power_law_1.2,0.32264320850372313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,4,power_law_1.2,2.778848075866699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,4,power_law_1.2,0.42418560981750486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,4,power_law_1.2,5.795603179931641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,4,power_law_1.2,0.5112576007843017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,4,power_law_1.2,0.6757120132446289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,4,power_law_1.2,0.8832320213317871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,4,power_law_1.2,1.2085184097290038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,4,power_law_1.2,1.5456192016601562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,4,power_law_1.2,1.9619199752807617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,4,power_law_1.2,2.859846305847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,4,power_law_1.2,6.1096446990966795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,2,balanced,0.06871466835339864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,4,power_law_1.2,0.07735679745674133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,2,balanced,0.07195733487606049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,2,balanced,0.08815999825795491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,4,power_law_1.2,0.09251199960708618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,2,balanced,0.11296000083287557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,2,balanced,0.14851199587186178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,4,power_law_1.2,0.09772160053253173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,2,balanced,0.22702399889628092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,4,power_law_1.2,0.08773760199546814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,2,balanced,0.3086400032043457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,2,balanced,0.3092319965362549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,4,power_law_1.2,0.1091647982597351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,2,balanced,0.3059733311335246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,2,balanced,0.3057493368784587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,4,power_law_1.2,0.13189120292663575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,2,balanced,0.30703999598821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,4,power_law_1.2,0.13868160247802735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,2,balanced,0.30612800518671673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,4,power_law_1.2,0.16618239879608154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,2,balanced,0.305786669254303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,2,balanced,0.3086719910303752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,4,power_law_1.2,0.17614079713821412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,2,balanced,0.31512532631556195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,4,power_law_1.2,0.1817728042602539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,2,balanced,0.32038400570551556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,4,power_law_1.2,0.1815551996231079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,2,balanced,0.32993600765864056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,4,power_law_1.2,0.18754559755325317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,2,balanced,0.3513866662979126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,4,power_law_1.2,0.18772480487823487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,2,balanced,0.3707893292109172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,4,power_law_1.2,0.19212160110473633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,4,power_law_1.2,0.20252161026000975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,2,balanced,0.4132426579793294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,4,power_law_1.2,0.21415040493011475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,4,power_law_1.2,0.2193727970123291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,2,balanced,0.4519573450088501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,4,power_law_1.2,0.23929600715637206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,4,power_law_1.2,0.2610111951828003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,2,balanced,0.5315146843592325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,4,power_law_1.2,0.29884159564971924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,4,power_law_1.2,0.3524032115936279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,4,power_law_1.2,0.42535037994384767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,2,balanced,0.6103359858194987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,4,power_law_1.2,0.5170623779296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,4,power_law_1.2,0.6911424160003662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,4,power_law_1.2,0.8812543869018554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,4,power_law_1.2,1.1719296455383301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,2,balanced,0.7843039830525717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,4,power_law_1.2,1.5440640449523926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,4,power_law_1.2,1.9322431564331055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,4,power_law_1.2,2.9226688385009765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,2,balanced,1.0647146701812744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,4,power_law_1.2,5.239487838745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,2,balanced,1.4072960217793782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,2,balanced,0.06278400123119354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,2,balanced,0.06537599861621857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,2,balanced,0.0721013347307841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,2,balanced,0.09461333354314168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,2,balanced,0.12311466534932454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,2,balanced,0.17847466468811035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,2,balanced,0.23053866624832153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,2,balanced,1.833877404530843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,2,balanced,0.23476799329121908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,2,balanced,0.23180800676345825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,2,balanced,0.23499733209609985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,2,balanced,0.23456533749898276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,2,balanced,0.2414026657740275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,2,balanced,0.08107733229796092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,2,balanced,0.08627733588218689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,2,balanced,0.24387200673421225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,2,balanced,0.0983840028444926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,2,balanced,0.24678399165471396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,2,balanced,0.12987732887268066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,2,balanced,0.16994667053222656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,2,balanced,0.25733333826065063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,2,balanced,0.2638559937477112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,2,balanced,0.2604479988416036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,2,balanced,0.3563093344370524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,2,balanced,0.3401333491007487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,2,balanced,0.2719786763191223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,2,balanced,2.2372800509134927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,2,balanced,0.35181868076324463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,2,balanced,0.2990079919497172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,2,balanced,0.33107733726501465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,2,balanced,0.3442080020904541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,2,balanced,0.3202293316523234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,2,balanced,0.3293600082397461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,2,balanced,0.3381439844767253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,2,balanced,0.3388693332672119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,2,balanced,0.3713066577911377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,2,balanced,0.35599998633066815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,2,balanced,0.3535573482513428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,2,balanced,0.41384534041086835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,2,balanced,0.36189866065979004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,2,balanced,0.381717324256897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,2,balanced,0.4086079994837443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,2,balanced,0.5077546834945679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,2,balanced,0.45266131560007733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,2,balanced,0.4901226758956909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,2,balanced,0.5987039804458618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,2,balanced,3.3943840662638345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,2,balanced,0.5847839911778768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,2,balanced,0.6737333138783773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,2,balanced,0.7978506882985433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,2,balanced,0.8575786749521891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,2,balanced,1.079925298690796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,2,balanced,1.2096479733784993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,2,balanced,1.451701323191325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,2,balanced,1.5690827369689941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,2,balanced,2.088319937388102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,2,balanced,1.915818691253662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,2,balanced,6.436607996622722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,2,power_law_1.01,0.07332479953765869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,2,power_law_1.01,0.08684160113334656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,2,power_law_1.01,0.10844800472259522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,2,power_law_1.01,0.10405759811401367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,2,power_law_1.01,0.1276479959487915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,2,power_law_1.01,0.163481605052948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,2,balanced,2.587183952331543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,2,power_law_1.01,0.1886080026626587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,2,power_law_1.01,0.22140800952911377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,2,power_law_1.01,0.22143359184265138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,2,power_law_1.01,0.2260416030883789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,2,power_law_1.01,0.2368000030517578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,2,power_law_1.01,0.23900799751281737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,2,balanced,2.376490592956543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,2,power_law_1.01,0.25064959526062014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,2,power_law_1.01,0.2619584083557129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,2,power_law_1.01,0.28168959617614747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,2,power_law_1.01,0.282694411277771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,2,power_law_1.01,0.3043904066085815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,2,power_law_1.01,0.32154240608215334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,2,power_law_1.01,0.3479743957519531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,2,power_law_1.01,0.40383358001708985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,2,power_law_1.01,0.44988160133361815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,2,power_law_1.01,0.5484032154083252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,2,balanced,3.980997403462728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,2,power_law_1.01,0.6581183910369873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,2,power_law_1.01,0.8593536376953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,2,power_law_1.01,1.0515328407287599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,2,power_law_1.01,1.4408448219299317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,2,power_law_1.01,1.8999744415283204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,2,power_law_1.01,2.315667152404785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,2,power_law_1.01,3.379359817504883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,2,balanced,3.6692212422688804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,2,power_law_1.01,6.942707061767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,2,balanced,7.683274586995442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,2,power_law_1.01,0.06465920209884643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,2,power_law_1.01,0.0726207971572876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,2,power_law_1.01,0.08698239922523499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,2,balanced,7.125200271606445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,2,power_law_1.01,0.0838271975517273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,2,power_law_1.01,0.10858240127563476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,2,power_law_1.01,0.13788800239562987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,2,power_law_1.01,0.14792319536209106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,2,power_law_1.01,0.174399995803833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,2,power_law_1.01,0.17965439558029175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,2,power_law_1.01,0.18581119775772095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,2,power_law_1.01,0.18899199962615967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,2,power_law_1.01,0.19950079917907715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,2,power_law_1.01,0.2011712074279785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,2,power_law_1.01,0.2142335891723633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,2,power_law_1.01,0.22939519882202147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,2,power_law_1.01,0.2415424108505249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,2,power_law_1.01,0.25021440982818605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,2,power_law_1.01,0.27663359642028806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,2,power_law_1.01,0.30714240074157717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,2,power_law_1.01,0.3631295919418335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,2,power_law_1.01,0.41621761322021483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,2,power_law_1.01,0.5296703815460205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,2,power_law_1.01,0.641651201248169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,2,power_law_1.01,0.8651904106140137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,2,power_law_1.01,1.0687040328979491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,2,power_law_1.01,1.5553536415100098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,2,power_law_1.01,1.9372095108032226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,2,power_law_1.01,2.4336896896362306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,2,power_law_1.01,3.8802558898925783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,2,power_law_1.01,0.08714879751205444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,2,power_law_1.01,7.274822235107422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,2,power_law_1.01,0.10088959932327271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,2,power_law_1.01,0.12167680263519287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,2,power_law_1.01,0.11640959978103638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,2,power_law_1.01,0.15281280279159545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,2,power_law_1.01,0.19201279878616334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,2,power_law_1.01,0.21088640689849852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,2,power_law_1.01,0.2388672113418579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,2,power_law_1.01,0.25867519378662107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,2,power_law_1.01,0.2579200029373169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,2,power_law_1.01,0.27078399658203123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,2,power_law_1.01,0.2714240074157715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,2,power_law_1.01,0.2784960031509399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,2,power_law_1.01,0.2848959922790527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,2,power_law_1.01,0.3095871925354004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,2,power_law_1.01,0.3089600086212158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,2,power_law_1.01,0.3244927883148193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,2,power_law_1.01,0.34650239944458006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,2,power_law_1.01,0.375980806350708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,2,power_law_1.01,0.43791999816894533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,2,power_law_1.01,0.500767993927002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,2,power_law_1.01,0.6175615787506104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,2,power_law_1.01,0.7412543773651123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,2,power_law_1.01,0.9675264358520508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,2,power_law_1.01,1.2470527648925782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,2,power_law_1.01,1.6904256820678711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,2,power_law_1.2,0.0717248022556305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,2,power_law_1.01,2.2410879135131836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,2,power_law_1.2,0.08579840064048767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,2,power_law_1.01,2.6879167556762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,2,power_law_1.2,0.09395840167999267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,2,power_law_1.01,4.172832107543945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,2,power_law_1.2,0.10250240564346313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,2,power_law_1.2,0.11853439807891845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,2,power_law_1.01,7.7871040344238285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,2,power_law_1.2,0.1492095947265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,2,power_law_1.2,0.17386239767074585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,2,power_law_1.2,0.20718719959259033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,2,power_law_1.2,0.21130239963531494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,2,power_law_1.2,0.22323200702667237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,2,power_law_1.2,0.2299328088760376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,2,power_law_1.2,0.23906559944152833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,2,power_law_1.2,0.24908800125122071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,2,power_law_1.2,0.26163198947906496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,2,power_law_1.2,0.06573439836502075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,2,power_law_1.2,0.2806272029876709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,2,power_law_1.2,0.07284479737281799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,2,power_law_1.2,0.2862080097198486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,2,power_law_1.2,0.07597439885139465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,2,power_law_1.2,0.3094912052154541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,2,power_law_1.2,0.08417279720306396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,2,power_law_1.2,0.33034238815307615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,2,power_law_1.2,0.1041152000427246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,2,power_law_1.2,0.36161279678344727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,2,power_law_1.2,0.12375680208206177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,2,power_law_1.2,0.41075839996337893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,2,power_law_1.2,0.1444607973098755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,2,power_law_1.2,0.16517119407653807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,2,power_law_1.2,0.4676032066345215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,2,power_law_1.2,0.1778880000114441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,2,power_law_1.2,0.5693183898925781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,2,power_law_1.2,0.17834880352020263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,2,power_law_1.2,0.6795904159545898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,2,power_law_1.2,0.18534400463104247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,2,power_law_1.2,0.8689536094665528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,2,power_law_1.2,0.19944319725036622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,2,power_law_1.2,1.1053312301635743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,2,power_law_1.2,0.2001215934753418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,2,power_law_1.2,1.5871616363525392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,2,power_law_1.2,0.21148159503936767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,2,power_law_1.2,1.9191104888916015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,2,power_law_1.2,0.22969601154327393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,2,power_law_1.2,0.23329920768737794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,2,power_law_1.2,2.2101055145263673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,2,power_law_1.2,0.2541248083114624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,2,power_law_1.2,3.61844482421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,2,power_law_1.2,0.278054404258728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,2,power_law_1.2,0.3108288049697876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,2,power_law_1.2,7.032441711425781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,2,power_law_1.2,0.37699840068817136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,2,power_law_1.2,0.42919039726257324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,2,power_law_1.2,0.5544832229614258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,2,power_law_1.2,0.6485375881195068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,2,power_law_1.2,0.8992639541625976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,2,power_law_1.2,0.08831359744071961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,2,power_law_1.2,1.1353471755981446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,2,power_law_1.2,0.09807360172271729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,2,power_law_1.2,1.5568575859069824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,2,power_law_1.2,0.09802240133285522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,2,power_law_1.2,2.021011161804199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,2,power_law_1.2,0.1141055941581726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,2,power_law_1.2,0.14256000518798828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,2,power_law_1.2,2.5574527740478517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,2,power_law_1.2,0.1754047989845276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,2,power_law_1.2,3.899308776855469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,2,power_law_1.2,0.19865599870681763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,2,power_law_1.2,0.24272639751434327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,2,power_law_1.2,7.574527740478516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,2,power_law_1.2,0.24944639205932617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,2,power_law_1.2,0.25559039115905763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,2,power_law_1.2,0.2630975961685181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,2,power_law_1.2,0.2790080070495605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,2,power_law_1.2,0.27854719161987307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,2,power_law_1.2,0.2883647918701172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,2,power_law_1.2,0.29675519466400146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,2,power_law_1.2,0.30346240997314455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,2,power_law_1.2,0.31943039894104003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,2,power_law_1.2,0.3515392065048218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,2,power_law_1.2,0.38611199855804446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,2,power_law_1.2,0.4515520095825195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,2,power_law_1.2,0.5151231765747071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,2,power_law_1.2,0.6349760055541992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,2,power_law_1.2,0.7654848098754883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,2,power_law_1.2,1.0305536270141602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,2,power_law_1.2,1.2540800094604492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,2,power_law_1.2,1.7261312484741211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,2,power_law_1.2,2.2645503997802736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,2,power_law_1.2,2.779078483581543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,2,power_law_1.2,4.32567024230957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,1,balanced,0.08298133313655853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,1,balanced,0.08629332979520161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,1,balanced,0.10032533605893452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,2,power_law_1.2,7.998905944824219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,1,balanced,0.1357866624991099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,1,balanced,0.17965332667032877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,1,balanced,0.2828106681505839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,1,balanced,0.38782934347788495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,1,balanced,0.39350934823354083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,1,balanced,0.39428265889485675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,1,balanced,0.3914719820022583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,1,balanced,0.396448016166687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,1,balanced,0.40415998299916583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,1,balanced,0.40697065989176434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,1,balanced,0.4131786823272705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,1,balanced,0.42235199610392254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,1,balanced,0.42483198642730713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,1,balanced,0.44070398807525635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,1,balanced,0.4851466814676921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,1,balanced,0.514304002126058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,1,balanced,0.5852266550064087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,1,balanced,0.6434933344523112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,1,balanced,0.7727786699930826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,1,balanced,0.8971733252207438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,1,balanced,1.1802666982014973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,1,balanced,0.07746133208274841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,1,balanced,0.07890133559703827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,1,balanced,0.09063999851544698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,1,balanced,0.12146133184432983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,1,balanced,1.6354773839314778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,1,balanced,0.17415465911229452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,1,balanced,0.2659466663996379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,1,balanced,0.35423465569814044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,1,balanced,0.35844798882802326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,1,balanced,0.35888532797495526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,1,balanced,0.3598986864089966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,1,balanced,0.3628053267796834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,1,balanced,0.3710026741027832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,1,balanced,0.37196799119313556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,1,balanced,0.3835039933522542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,1,balanced,0.38866134484608966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,1,balanced,2.1916747093200684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,1,balanced,0.3947519858678182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,1,balanced,0.4121706485748291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,1,power_law_1.01,0.07386879920959473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,1,balanced,0.4559359947840373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,1,power_law_1.01,0.08135039806365967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,1,balanced,0.48737601439158124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,1,power_law_1.01,0.09524480104446412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,1,power_law_1.01,0.11804159879684448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,1,balanced,0.5682559808095297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,1,power_law_1.01,0.15576319694519042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,1,power_law_1.01,0.18867839574813844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,1,balanced,0.6335200071334839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,1,power_law_1.01,0.2257920026779175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,1,power_law_1.01,0.26661760807037355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,1,balanced,0.7815093199412028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,1,power_law_1.01,0.2822720050811768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,1,balanced,2.8776639302571616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,1,power_law_1.01,0.2843008041381836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,1,power_law_1.01,0.2999232053756714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,1,balanced,0.9236853122711182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,1,power_law_1.01,0.31351680755615235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,1,power_law_1.01,0.3187455892562866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,1,power_law_1.01,0.3339904069900513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,1,power_law_1.01,0.34729599952697754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,1,balanced,1.2133546670277913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,1,power_law_1.01,0.3497535943984985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,1,power_law_1.01,0.37954559326171877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,1,power_law_1.01,0.44716801643371584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,1,power_law_1.01,0.4764224052429199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,1,balanced,1.7015573183695476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,1,power_law_1.01,0.5642623901367188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,1,power_law_1.01,0.6306367874145508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,1,balanced,3.580575942993164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,1,power_law_1.01,0.7947264194488526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,1,power_law_1.01,0.9482111930847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,1,power_law_1.01,1.2622079849243164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,1,power_law_1.01,1.5685824394226073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,1,power_law_1.01,0.06752640008926392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,1,balanced,2.2863945960998535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,1,power_law_1.01,2.188800048828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,1,power_law_1.01,0.07501440048217774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,1,power_law_1.01,0.08803840279579163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,1,power_law_1.01,2.8312320709228516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,1,power_law_1.01,0.11054079532623291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,1,power_law_1.01,3.480838394165039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,1,power_law_1.01,0.14295680522918702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,1,power_law_1.01,5.367225646972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,1,power_law_1.01,0.17889280319213868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,1,power_law_1.01,0.21176960468292236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,1,power_law_1.01,10.422035217285156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,1,power_law_1.01,0.24744958877563478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,1,power_law_1.01,0.2623424053192139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,1,balanced,3.064133326212565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,1,power_law_1.01,0.27160959243774413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,1,power_law_1.01,0.2796799898147583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,1,power_law_1.01,0.2944000005722046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,1,power_law_1.01,0.2980736017227173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,1,power_law_1.01,0.31169281005859373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,1,balanced,5.492240269978841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,1,power_law_1.01,0.3263423919677734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,1,power_law_1.01,0.34491519927978515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,1,power_law_1.01,0.3570559978485107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,1,power_law_1.01,0.41249918937683105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,1,power_law_1.01,0.44678401947021484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,1,balanced,3.8223466873168945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,1,power_law_1.01,0.5428287982940674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,1,power_law_1.01,0.6221312046051025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,1,power_law_1.01,0.8012096405029296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,1,power_law_1.01,0.9751999855041504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,1,power_law_1.01,1.3221119880676269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,1,power_law_1.01,1.6641984939575196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,1,power_law_1.01,2.347987174987793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,1,power_law_1.01,3.0468095779418944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,1,power_law_1.01,3.7319614410400392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,1,power_law_1.01,5.8064830780029295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,1,power_law_1.01,11.369574737548827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,1,balanced,5.901519775390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,1,balanced,10.559893290201822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,1,balanced,11.46133804321289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,1,power_law_1.2,0.0741823971271515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,1,power_law_1.2,0.08074880242347718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,1,power_law_1.2,0.09175040125846863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,1,power_law_1.2,0.11611520051956177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,1,power_law_1.2,0.14485119581222533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,1,power_law_1.2,0.17152639627456664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,1,power_law_1.2,0.20373759269714356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,1,power_law_1.2,0.25613439083099365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,1,power_law_1.2,0.2695552110671997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,1,power_law_1.2,0.28234879970550536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,1,power_law_1.2,0.2927680015563965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,1,power_law_1.2,0.309004807472229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,1,power_law_1.2,0.31484160423278806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,1,power_law_1.2,0.3354880094528198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,8,balanced,0.05063466727733612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,1,power_law_1.2,0.34506878852844236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,8,balanced,0.051226665576299034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,8,balanced,0.04924799998601278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,1,power_law_1.2,0.35683839321136473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,8,balanced,0.050986667474110924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,1,power_law_1.2,0.3930176019668579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,8,balanced,0.0528106689453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,8,balanced,0.052330667773882546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,1,power_law_1.2,0.06758400201797485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,1,power_law_1.2,0.4481088161468506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,8,balanced,0.05043200155099233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,1,power_law_1.2,0.07398399710655212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,8,balanced,0.04850666721661886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,1,power_law_1.2,0.49431681632995605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,8,balanced,0.05053333441416422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,1,power_law_1.2,0.08417279720306396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,1,power_law_1.2,0.574828815460205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,8,balanced,0.0505920002857844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,1,power_law_1.2,0.10965759754180908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,8,balanced,0.04882133503754934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,1,power_law_1.2,0.6454783916473389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,8,balanced,0.04668800036112467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,1,power_law_1.2,0.1368384003639221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,1,power_law_1.2,0.8177215576171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,8,balanced,0.044938668608665466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,8,balanced,0.04663999875386556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,1,power_law_1.2,0.16619520187377929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,1,power_law_1.2,0.9629247665405274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,8,balanced,0.04850133260091146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,1,power_law_1.2,0.18855040073394774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,1,power_law_1.2,1.2832063674926757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,8,balanced,0.04641066491603851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,1,power_law_1.2,0.24369919300079346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,8,balanced,0.050245334704717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,1,power_law_1.2,1.5917887687683105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,8,balanced,0.053344001372655235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,1,power_law_1.2,0.25516159534454347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,1,power_law_1.2,2.2215871810913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,8,balanced,0.054042667150497437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,1,power_law_1.2,0.2603967905044556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,8,balanced,0.058746665716171265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,1,power_law_1.2,2.843328094482422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,1,power_law_1.2,0.2709120035171509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,8,balanced,0.06252799928188324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,1,power_law_1.2,0.284716796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,1,power_law_1.2,3.4846527099609377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,8,balanced,0.08698667089144389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,1,power_law_1.2,0.2981503963470459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,1,power_law_1.2,5.370969772338867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,8,balanced,0.09219200412432353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,1,power_law_1.2,0.31232640743255613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,8,balanced,0.040021332601706185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,8,balanced,0.04077333211898804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,1,power_law_1.2,0.3236991882324219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,8,balanced,0.12210667133331299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,1,power_law_1.2,10.432870483398437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,8,balanced,0.04071466624736786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,8,balanced,0.04213866591453552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,1,power_law_1.2,0.3368256092071533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,8,balanced,0.04117333392302195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,1,power_law_1.2,0.3646591901779175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,8,balanced,0.13885333140691122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,8,balanced,0.043247997760772705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,8,balanced,0.04484266539414724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,1,power_law_1.2,0.42388482093811036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,8,balanced,0.04283200204372406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,1,power_law_1.2,0.4677055835723877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,8,balanced,0.043280000487963356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,8,balanced,0.19509865840276083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,8,balanced,0.043578664461771645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,1,power_law_1.2,0.5627071857452393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,8,balanced,0.04298133154710134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,8,balanced,0.042778665820757546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,1,power_law_1.2,0.6376704216003418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,8,balanced,0.04248000184694926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,1,power_law_1.2,0.8229311943054199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,8,balanced,0.2371413310368856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,8,balanced,0.04265599946180979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,8,balanced,0.04641599953174591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,1,power_law_1.2,0.9956735610961914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,8,balanced,0.04747200012207031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,8,balanced,0.04776533444722494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,1,power_law_1.2,1.3361087799072267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,8,balanced,0.053455998500188194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,8,balanced,0.294869323571523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,1,power_law_1.2,1.679520034790039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,8,balanced,0.055359999338785805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,1,power_law_1.2,2.3689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,8,balanced,0.05852266649405161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,8,balanced,0.0670826683441798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,1,power_law_1.2,3.0474496841430665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,8,balanced,0.09671466549237569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,1,power_law_1.2,3.74389762878418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,8,balanced,0.4368640184402466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,8,balanced,0.11689600348472595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,1,power_law_1.2,5.818796920776367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,8,balanced,0.15597866972287497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,1,power_law_1.2,11.36895980834961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,8,balanced,0.18745599190394083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,8,balanced,0.26104533672332764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,8,balanced,0.8345226446787516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,8,balanced,0.3273973266283671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,8,balanced,0.4021386702855428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,8,balanced,0.6142986615498861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,8,balanced,1.1798986593882244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,8,balanced,0.04026666780312856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,8,balanced,0.041434665520985924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,8,balanced,0.040906667709350586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,8,balanced,0.04338666796684265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,8,balanced,0.04276266694068909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,8,balanced,0.04313066601753235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,8,balanced,0.04497600098450979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,8,balanced,0.04341333111127218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,8,balanced,0.04292800029118856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,8,balanced,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,8,balanced,0.04372799893220266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,8,balanced,0.04341333111127218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,8,balanced,0.04248000184694926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,8,balanced,0.04281599819660187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,8,balanced,0.04660800099372864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,8,balanced,0.045253331462542214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,8,balanced,0.046725332736968994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,8,balanced,0.051226665576299034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,8,balanced,0.054229333996772766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,8,balanced,0.060991997520128884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,8,balanced,0.019472000499566395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,8,balanced,0.06540800134340923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,8,balanced,0.020608000457286835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,8,balanced,0.020256000260512035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,8,balanced,0.08161599934101105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,8,balanced,0.021029333273569744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,8,balanced,0.02143999934196472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,8,balanced,0.022874665757020313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,8,balanced,0.09284266829490662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,8,balanced,0.04048533240954081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,8,balanced,0.04109866668780645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,8,balanced,0.11899200081825256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,8,balanced,0.033002667129039764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,8,power_law_1.01,0.049132800102233885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,8,balanced,0.03245333333810171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,8,power_law_1.01,0.04822399914264679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,8,balanced,0.03209066639343897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,8,balanced,0.14005333185195923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,8,power_law_1.01,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,8,balanced,0.03721600025892258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,8,balanced,0.03446399917205175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,8,balanced,0.034261333445707955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,8,balanced,0.040549332896868386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,8,balanced,0.19019200404485068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,8,balanced,0.03458133339881897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,8,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,8,balanced,0.04925866425037384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,8,balanced,0.051088000337282814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,8,balanced,0.06453866759936015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,8,balanced,0.23760000864664713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,8,balanced,0.07049599786599477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,8,balanced,0.09724266330401103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,8,balanced,0.11339199542999268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,8,balanced,0.2844480077425639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,8,balanced,0.15747732917467752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,8,balanced,0.19854400555292764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,8,balanced,0.28436799844106037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,8,balanced,0.42904531955718994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,8,balanced,0.36877334117889404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,8,power_law_1.01,0.03829759955406189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,8,power_law_1.01,0.0392192006111145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,8,power_law_1.01,0.037913599610328676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,8,power_law_1.01,0.03983359932899475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,8,power_law_1.01,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,8,power_law_1.01,0.040454399585723874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,8,balanced,0.45360000928243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,8,power_law_1.01,0.04035199880599975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,8,power_law_1.01,0.039743998646736146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,8,power_law_1.01,0.03937279880046844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,8,power_law_1.01,0.04020479917526245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,8,power_law_1.01,0.039987200498580934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,8,power_law_1.01,0.04147199988365173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,8,power_law_1.01,0.04310399889945984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,8,balanced,0.7067946592966715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,8,power_law_1.01,0.042668798565864564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,8,balanced,0.8059786955515543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,8,power_law_1.01,0.04769279956817627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,8,power_law_1.01,0.04915199875831604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,8,power_law_1.01,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,8,power_law_1.01,0.06284160017967225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,8,power_law_1.01,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,8,power_law_1.01,0.08915200233459472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,8,power_law_1.01,0.10798720121383668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,8,power_law_1.01,0.15173120498657228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,8,power_law_1.01,0.17744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,8,power_law_1.01,0.24419839382171632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,8,balanced,1.3826826413472493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,8,power_law_1.01,0.3071104049682617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,8,power_law_1.01,0.44832639694213866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,8,power_law_1.01,0.6028096199035644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,8,power_law_1.01,0.7507264137268066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,8,power_law_1.01,1.193619155883789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,8,power_law_1.01,2.357529640197754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,8,power_law_1.01,0.038624000549316403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,8,power_law_1.01,0.04083200097084046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,8,power_law_1.01,0.038700801134109494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,8,power_law_1.01,0.039628800749778745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,8,power_law_1.01,0.04142720103263855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,8,power_law_1.01,0.04033919870853424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,8,power_law_1.01,0.0404992014169693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,8,power_law_1.01,0.040524798631668094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,8,power_law_1.01,0.018406400084495546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,8,power_law_1.01,0.04088320136070252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,8,power_law_1.01,0.019577600061893463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,8,power_law_1.01,0.03880319893360138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,8,power_law_1.01,0.0190080001950264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,8,power_law_1.01,0.039084801077842714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,8,power_law_1.01,0.01883520036935806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,8,power_law_1.01,0.04211199879646301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,8,power_law_1.01,0.019468800723552705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,8,power_law_1.01,0.043532800674438474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,8,power_law_1.01,0.04776960015296936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,8,power_law_1.01,0.02200320065021515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,8,power_law_1.01,0.04273279905319214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,8,power_law_1.01,0.03834879994392395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,8,power_law_1.01,0.04922240078449249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,8,power_law_1.01,0.04880000054836273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,8,power_law_1.01,0.050329601764678954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,8,power_law_1.01,0.04713599979877472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,8,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,8,power_law_1.01,0.031116798520088196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,8,power_law_1.01,0.054816001653671266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,8,power_law_1.01,0.046265599131584165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,8,power_law_1.2,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,8,power_law_1.01,0.03079040050506592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,8,power_law_1.01,0.060819202661514284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,8,power_law_1.01,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,8,power_law_1.2,0.048172798752784726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,8,power_law_1.01,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.01,0.06784639954566955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,8,power_law_1.01,0.04403199851512909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,8,power_law_1.2,0.04888319969177246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,8,power_law_1.01,0.03526400029659271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.01,0.08171520233154297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,8,power_law_1.01,0.044249600172042845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,8,power_law_1.2,0.0494271993637085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,8,power_law_1.01,0.034297600388526917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.01,0.09182080030441284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,8,power_law_1.01,0.043654400110244754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,8,power_law_1.01,0.032416000962257385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.01,0.1238592028617859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,8,power_law_1.01,0.04830079972743988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.01,0.15253119468688964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,8,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,8,power_law_1.01,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,8,power_law_1.01,0.0339711993932724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,8,power_law_1.01,0.047219198942184445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.01,0.21001598834991456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,8,power_law_1.01,0.03560320138931274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,8,power_law_1.01,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.01,0.25620479583740235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,8,power_law_1.01,0.046623998880386354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,8,power_law_1.01,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.01,0.388153600692749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,8,power_law_1.01,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.01,0.048537600040435794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.01,0.4963200092315674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,8,power_law_1.01,0.07637119889259339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.01,0.06306560039520263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,8,power_law_1.01,0.6181439876556396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,8,power_law_1.01,0.08117759823799134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.01,0.06941440105438232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,8,power_law_1.01,0.8647168159484864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,8,power_law_1.01,0.09997439980506898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.01,0.09593600034713745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,8,power_law_1.01,1.707916831970215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,8,power_law_1.01,0.11545599699020385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.01,0.11267839670181275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,8,power_law_1.01,0.1522304058074951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.01,0.157043194770813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,8,power_law_1.01,0.1929535984992981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.01,0.19879679679870604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,8,power_law_1.01,0.2753727912902832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.01,0.284550404548645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,8,power_law_1.01,0.3436543941497803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.01,0.36830720901489256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,8,power_law_1.01,0.5146880149841309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,8,power_law_1.01,0.45380477905273436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,8,power_law_1.01,0.6508416175842285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,8,power_law_1.01,0.7082496166229248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,8,power_law_1.01,0.8417152404785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,8,power_law_1.01,1.383232021331787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,8,power_law_1.01,1.3621439933776855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,8,power_law_1.01,2.6915519714355467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,8,power_law_1.2,0.038815999031066896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,8,power_law_1.2,0.03829120099544525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,8,power_law_1.2,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,8,power_law_1.2,0.040243199467658995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,8,power_law_1.2,0.0401856005191803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,8,power_law_1.2,0.04097279906272888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,8,power_law_1.2,0.018137599527835845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,8,power_law_1.2,0.040115201473236085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,8,power_law_1.2,0.03933440148830414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,8,power_law_1.2,0.018060800433158875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,8,power_law_1.2,0.039315199851989745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,8,power_law_1.2,0.03927040100097656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,8,power_law_1.2,0.01820160001516342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,8,power_law_1.2,0.03960959911346436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,8,power_law_1.2,0.04044159948825836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,8,power_law_1.2,0.019251200556755065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,8,power_law_1.2,0.039955198764801025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,8,power_law_1.2,0.04136320054531097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,8,power_law_1.2,0.021568000316619873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,8,power_law_1.2,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,8,power_law_1.2,0.041740798950195314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,8,power_law_1.2,0.021478399634361267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,8,power_law_1.2,0.04529919922351837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,8,power_law_1.2,0.0411327987909317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,8,power_law_1.2,0.03839359879493713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,8,power_law_1.2,0.04165120124816894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,8,power_law_1.2,0.04103679955005646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,8,power_law_1.2,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,8,power_law_1.2,0.0440064013004303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,8,power_law_1.2,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,8,power_law_1.2,0.03225600123405457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,8,power_law_1.2,0.049798399209976196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,8,power_law_1.2,0.040889599919319154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,8,power_law_1.2,0.03025279939174652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,8,power_law_1.2,0.05004159808158874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,8,power_law_1.2,0.03925760090351105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,8,power_law_1.2,0.03091199994087219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,8,power_law_1.2,0.052198398113250735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,8,power_law_1.2,0.03909119963645935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,8,power_law_1.2,0.035206401348114015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,8,power_law_1.2,0.06466559767723083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,8,power_law_1.2,0.04239999949932098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,8,power_law_1.2,0.03529599905014038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,8,power_law_1.2,0.07201279997825623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,8,power_law_1.2,0.04500479996204376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,8,power_law_1.2,0.033983999490737916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,8,power_law_1.2,0.08970879912376403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,8,power_law_1.2,0.04356479942798615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,8,power_law_1.2,0.03710080087184906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,8,power_law_1.2,0.04983679950237274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,8,power_law_1.2,0.107315194606781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,8,power_law_1.2,0.05201280117034912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,8,power_law_1.2,0.03338240087032318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,8,power_law_1.2,0.15273599624633788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,8,power_law_1.2,0.05425919890403748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,8,power_law_1.2,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,8,power_law_1.2,0.17783679962158203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,8,power_law_1.2,0.06035839915275574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,8,power_law_1.2,0.046003198623657225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,8,power_law_1.2,0.24804480075836183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.2,0.06716160178184509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.2,0.04806399941444397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,8,power_law_1.2,0.3333823919296265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.2,0.08060160279273987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.2,0.06119040250778198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,8,power_law_1.2,0.48051838874816893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.2,0.08963840007781983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.2,0.06892160177230836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,8,power_law_1.2,0.6259647846221924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.2,0.12480000257492066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.2,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,8,power_law_1.2,0.7776256084442139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.2,0.11147520542144776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.2,0.15581439733505248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,8,power_law_1.2,1.2126527786254884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.2,0.1558527946472168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.2,0.2134335994720459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,8,power_law_1.2,2.3547008514404295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.2,0.2599040031433105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.2,0.1977023959159851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.2,0.3489408016204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.2,0.28453760147094725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.2,0.500383996963501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.2,0.36951680183410646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,8,power_law_1.2,0.593721580505371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,8,power_law_1.2,0.4527423858642578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,8,power_law_1.2,0.7056320190429688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,8,power_law_1.2,0.9600255966186524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,8,power_law_1.2,1.3808704376220704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,8,power_law_1.2,1.7734783172607422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,8,power_law_1.2,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,8,power_law_1.2,0.04909439980983734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,8,power_law_1.2,0.047712001204490664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,8,power_law_1.2,0.045952001214027406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,8,power_law_1.2,0.04708479940891266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,8,power_law_1.2,0.0440064013004303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,8,power_law_1.2,0.044223999977111815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,8,power_law_1.2,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,8,power_law_1.2,0.04824320077896118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,8,power_law_1.2,0.047712001204490664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,8,power_law_1.2,0.0523904025554657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,8,power_law_1.2,0.0516543984413147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,8,power_law_1.2,0.060652798414230345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,8,power_law_1.2,0.07511039972305297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,8,power_law_1.2,0.08253440260887146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,8,power_law_1.2,0.10154240131378174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,8,power_law_1.2,0.1185215950012207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,8,power_law_1.2,0.15968639850616456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,8,power_law_1.2,0.18891520500183107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,8,power_law_1.2,0.25520639419555663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,8,power_law_1.2,0.35820159912109373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,8,power_law_1.2,0.5047679901123047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,8,power_law_1.2,0.6650815963745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,8,power_law_1.2,0.8536767959594727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,8,power_law_1.2,1.361900806427002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,8,power_law_1.2,2.6785856246948243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,4,balanced,0.04197866717974345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,4,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,4,balanced,0.04456533491611481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,4,balanced,0.045194665590922035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,4,balanced,0.04457599918047587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,4,balanced,0.046944002310434975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,4,balanced,0.035930665830771126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,4,balanced,0.03613866617282232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,4,balanced,0.04516266783078512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,4,balanced,0.03706666578849157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,4,balanced,0.03707200040419897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,4,balanced,0.04730133215586344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,4,balanced,0.04274133344491323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,4,balanced,0.04080000023047129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,4,balanced,0.04713066418965658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,4,balanced,0.041034666200478874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,4,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,4,balanced,0.04334933559099833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,4,balanced,0.045226668318112694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,4,balanced,0.04102933406829834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,4,balanced,0.04269866645336151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,4,balanced,0.04535999894142151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,4,balanced,0.04101866732041041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,4,balanced,0.0450186679760615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,4,balanced,0.048101335763931274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,4,balanced,0.04279466470082601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,4,balanced,0.04275199770927429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,4,balanced,0.0446720023949941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,4,balanced,0.04182933270931244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,4,balanced,0.04307733476161957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,4,balanced,0.045423999428749084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,4,balanced,0.042954668402671814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,4,balanced,0.04523199796676636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,4,balanced,0.04929600159327189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,4,balanced,0.040949332217375435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,4,balanced,0.04276266694068909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,4,balanced,0.047242666284243263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,4,balanced,0.042021334171295166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,4,balanced,0.042954668402671814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,4,balanced,0.04902400076389313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,4,balanced,0.04268266757329305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,4,balanced,0.044826666514078774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,4,balanced,0.044341335693995156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,4,balanced,0.054799998799959816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,4,balanced,0.04284266630808512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,4,balanced,0.04271466533342997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,4,balanced,0.05758399764696757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,4,balanced,0.04781866570313772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,4,balanced,0.0450186679760615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,4,balanced,0.06450133522351582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,4,balanced,0.046575998266537987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,4,balanced,0.04660800099372864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,4,balanced,0.04886933167775472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,4,balanced,0.04862933357556661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,4,balanced,0.07060266534487407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,4,balanced,0.05505066613356272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,4,balanced,0.05438933273156484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,4,balanced,0.0939573347568512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,4,balanced,0.056314667065938316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,4,balanced,0.0551093320051829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,4,balanced,0.06691733499368031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,4,balanced,0.06698666512966156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,4,balanced,0.10148800412813823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,4,balanced,0.07094933092594147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,4,balanced,0.07258133093516032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,4,balanced,0.13520532846450806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,4,balanced,0.08800533413887024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,4,balanced,0.10622933506965637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,4,balanced,0.10156800349553426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,4,balanced,0.12315199772516887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,4,balanced,0.019509332875410717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,4,balanced,0.15414933363596597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,4,balanced,0.020560000091791153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,4,balanced,0.02048533285657565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,4,balanced,0.13589333494504294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,4,balanced,0.16878400246302286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,4,balanced,0.020773333807786305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,4,balanced,0.020175999651352566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,4,balanced,0.21658132473627725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,4,balanced,0.163482666015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,4,balanced,0.020879998803138733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,4,balanced,0.20414932568868002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,4,balanced,0.024154665569464367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,4,balanced,0.023999998966852825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,4,balanced,0.04083733260631561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,4,balanced,0.22614934047063193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,4,balanced,0.29041600227355957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,4,balanced,0.04062933226426443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,4,balanced,0.2662186622619629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,4,balanced,0.04067199925581614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,4,balanced,0.03416533271471659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,4,balanced,0.03310933212439219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,4,balanced,0.2836906711260478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,4,balanced,0.03219199925661087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,4,balanced,0.36901334921518963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,4,balanced,0.036933332681655884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,4,balanced,0.034847999612490334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,4,balanced,0.3285333315531413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,4,balanced,0.03628266602754593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,4,balanced,0.041946664452552795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,4,balanced,0.3450506528218587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,4,balanced,0.44911468029022217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,4,balanced,0.04329599936803182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,4,balanced,0.05468800167242686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,4,balanced,0.06090133388837179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,4,balanced,0.07890133559703827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,4,balanced,0.49194133281707764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,4,balanced,0.5264159838358561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,4,balanced,0.6942880153656006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,4,balanced,0.08948799967765808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,4,balanced,0.12411200006802876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,4,balanced,0.1511573294798533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,4,balanced,0.21183999379475912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,4,balanced,1.0102720260620117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,4,balanced,1.3392853736877441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,4,balanced,0.27270400524139404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,4,balanced,0.9433173338572184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,4,balanced,0.3349066575368245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,4,balanced,0.5199893315633138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,4,balanced,1.0055359999338787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,4,power_law_1.01,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,4,power_law_1.01,0.04446719884872437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,4,power_law_1.01,0.03591679930686951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,4,power_law_1.01,0.04168320000171662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,4,power_law_1.01,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,4,power_law_1.01,0.04407680034637451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,4,power_law_1.01,0.043968001008033754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,4,power_law_1.01,0.03635199964046478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,4,power_law_1.01,0.045414400100708005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,4,power_law_1.01,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,4,power_law_1.01,0.04501760005950928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,4,power_law_1.01,0.03904640078544617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,4,power_law_1.01,0.04542079865932465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,4,power_law_1.01,0.039699199795722964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,4,power_law_1.01,0.0457152009010315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,4,power_law_1.01,0.041228801012039185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,4,power_law_1.01,0.04467839896678925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,4,power_law_1.01,0.03795199990272522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,4,power_law_1.01,0.040755200386047366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,4,power_law_1.01,0.044607999920845035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,4,power_law_1.01,0.03800959885120392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,4,power_law_1.01,0.04102399945259094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,4,power_law_1.01,0.044870400428771974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,4,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,4,power_law_1.01,0.04131839871406555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,4,power_law_1.01,0.04374400079250336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,4,power_law_1.01,0.04254719913005829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,4,power_law_1.01,0.041971200704574586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,4,power_law_1.01,0.046086400747299194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,4,power_law_1.01,0.04105600118637085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,4,power_law_1.01,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,4,power_law_1.01,0.0409280002117157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,4,power_law_1.01,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,4,power_law_1.01,0.04318720102310181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,4,power_law_1.01,0.04154880046844482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,4,power_law_1.01,0.055827200412750244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,4,power_law_1.01,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,4,power_law_1.01,0.04152320027351379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,4,power_law_1.01,0.05051519870758057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,4,power_law_1.01,0.042847999930381776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,4,power_law_1.01,0.05121279954910278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,4,power_law_1.01,0.04156799912452698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,4,power_law_1.01,0.05583360195159912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,4,power_law_1.01,0.04202240109443665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,4,power_law_1.01,0.058182400465011594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,4,power_law_1.01,0.043347200751304625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,4,power_law_1.01,0.06543999910354614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,4,power_law_1.01,0.04429439902305603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,4,power_law_1.01,0.08759679794311523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,4,power_law_1.01,0.045228800177574156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,4,power_law_1.01,0.10022399425506592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,4,power_law_1.01,0.05066239833831787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,4,power_law_1.01,0.13432320356369018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,4,power_law_1.01,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,4,power_law_1.01,0.17517440319061278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,4,power_law_1.01,0.05549439787864685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,4,power_law_1.01,0.2388159990310669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,4,power_law_1.01,0.017983999848365784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,4,power_law_1.01,0.06219519972801209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,4,power_law_1.01,0.3014847993850708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.01,0.06674559712409973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,4,power_law_1.01,0.01913599967956543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,4,power_law_1.01,0.45678081512451174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.01,0.0824895977973938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,4,power_law_1.01,0.018982400000095368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,4,power_law_1.01,0.649126386642456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.01,0.10213119983673095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,4,power_law_1.01,0.7518464088439941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.01,0.1263360023498535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.01,0.16666239500045776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,4,power_law_1.01,1.249292755126953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.01,0.22172799110412597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,4,power_law_1.01,2.4260992050170898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.01,0.2872191905975342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.01,0.4007999897003174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.01,0.5654335975646972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,4,power_law_1.01,0.6464640140533447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,4,power_law_1.01,1.011616039276123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,4,power_law_1.01,2.0962175369262694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,4,power_law_1.2,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,4,power_law_1.2,0.036473599076271054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,4,power_law_1.2,0.04131200015544891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,4,power_law_1.2,0.03635840117931366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,4,power_law_1.2,0.042387199401855466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,4,power_law_1.2,0.04145280122756958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,4,power_law_1.2,0.039577600359916684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,4,power_law_1.2,0.044915199279785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,4,power_law_1.2,0.03993600010871887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,4,power_law_1.2,0.0443583995103836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,4,power_law_1.2,0.039001598954200745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,4,power_law_1.2,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,4,power_law_1.2,0.04069760143756866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,4,power_law_1.2,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,4,power_law_1.2,0.0406464010477066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,4,power_law_1.2,0.04534400105476379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,4,power_law_1.2,0.04079999923706055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,4,power_law_1.2,0.044095999002456664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,4,power_law_1.2,0.04136320054531097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,4,power_law_1.2,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,4,power_law_1.2,0.040268799662590025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,4,power_law_1.2,0.04454399943351746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,4,power_law_1.2,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,4,power_law_1.2,0.045516800880432126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,4,power_law_1.2,0.04181120097637177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,4,power_law_1.2,0.0456959992647171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,4,power_law_1.2,0.04392960071563721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,4,power_law_1.2,0.04904960095882416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,4,power_law_1.2,0.04962559938430786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,4,power_law_1.2,0.05237759947776795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,4,power_law_1.2,0.04936319887638092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,4,power_law_1.2,0.05623679757118225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,4,power_law_1.2,0.0546239972114563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,4,power_law_1.2,0.06305919885635376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,4,power_law_1.2,0.06223999857902527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,4,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,4,power_law_1.2,0.06984959840774536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,4,power_law_1.2,0.085452800989151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,4,power_law_1.2,0.09253759980201721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,4,power_law_1.2,0.09357439875602722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,4,power_law_1.2,0.11596159934997559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,4,power_law_1.2,0.12301440238952636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,4,power_law_1.2,0.14970240592956544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,4,power_law_1.2,0.1448639988899231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,4,power_law_1.2,0.17445759773254393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,4,power_law_1.2,0.18495999574661254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,4,power_law_1.2,0.27515521049499514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,4,power_law_1.2,0.2873663902282715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,4,power_law_1.2,0.3454400062561035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,4,power_law_1.2,0.32337279319763185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,4,power_law_1.2,0.45557122230529784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,4,power_law_1.2,0.48458237648010255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,4,power_law_1.2,0.5987584114074707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,4,power_law_1.2,0.6625088214874267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,4,power_law_1.2,0.8302911758422852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,4,power_law_1.2,0.812992000579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,4,power_law_1.2,1.3422783851623534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,4,power_law_1.2,1.2287551879882812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,4,power_law_1.2,2.5083648681640627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,4,power_law_1.2,2.515763282775879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,4,power_law_1.01,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,4,power_law_1.01,0.06654719710350036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,4,power_law_1.01,0.07274879813194275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,4,power_law_1.01,0.10067839622497558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,4,power_law_1.01,0.11390719413757325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,4,power_law_1.01,0.1364416003227234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,4,power_law_1.01,0.16684160232543946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,4,power_law_1.01,0.25106561183929443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,4,power_law_1.01,0.29006080627441405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,4,power_law_1.01,0.4318784236907959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,4,power_law_1.01,0.5445119857788085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,4,power_law_1.01,0.7704448223114013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,4,power_law_1.01,1.0372608184814454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,4,power_law_1.01,0.018163199722766876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,4,power_law_1.01,0.019340799748897554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,4,power_law_1.01,2.044812774658203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,4,power_law_1.01,0.020006400346755982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,4,power_law_1.01,0.022150400280952453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,4,power_law_1.01,0.023039999604225158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,4,power_law_1.01,0.040755200386047366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,4,power_law_1.01,0.040870401263237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,4,power_law_1.01,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,4,power_law_1.01,0.033523198962211606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,4,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,4,power_law_1.01,0.03219839930534363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,4,power_law_1.01,0.03495680093765259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,4,power_law_1.01,0.03377920091152191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,4,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,4,power_law_1.01,0.03852800130844116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.01,0.04217599928379059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.01,0.05430399775505066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.01,0.060288000106811526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.01,0.0781503975391388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.01,0.08928639888763427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.01,0.1233407974243164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.01,0.15048320293426515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.01,0.21249918937683104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.01,0.2717632055282593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,4,power_law_1.01,0.3370879888534546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,4,power_law_1.01,0.5202688217163086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,4,power_law_1.01,1.0057472229003905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,4,power_law_1.2,0.037964800000190736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,4,power_law_1.2,0.01788160055875778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,4,power_law_1.2,0.04041599929332733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,4,power_law_1.2,0.019142399728298187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,4,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,4,power_law_1.2,0.01953279972076416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,4,power_law_1.2,0.04200319945812225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,4,power_law_1.2,0.019174399971961974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,4,power_law_1.2,0.0421887993812561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,4,power_law_1.2,0.019392000138759614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,4,power_law_1.2,0.042777600884437564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,4,power_law_1.2,0.01992959976196289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,4,power_law_1.2,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,4,power_law_1.2,0.02290560007095337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,4,power_law_1.2,0.04248960018157959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,4,power_law_1.2,0.022944000363349915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,4,power_law_1.2,0.04229120016098022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,4,power_law_1.2,0.039526399970054624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,4,power_law_1.2,0.04199039936065674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,4,power_law_1.2,0.0425024002790451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,4,power_law_1.2,0.04013440012931824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,4,power_law_1.2,0.04381439983844757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,4,power_law_1.2,0.043731200695037845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,4,power_law_1.2,0.04088320136070252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,4,power_law_1.2,0.04500479996204376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,4,power_law_1.2,0.032262399792671204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,2,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,4,power_law_1.2,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,2,balanced,0.04127466678619385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,4,power_law_1.2,0.03516159951686859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,2,balanced,0.04474666714668274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,4,power_law_1.2,0.052671998739242554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,4,power_law_1.2,0.03346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,2,balanced,0.04497600098450979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,4,power_law_1.2,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,4,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,2,balanced,0.04712533454100291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,4,power_law_1.2,0.0650816023349762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,4,power_law_1.2,0.035622400045394895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,2,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.2,0.0731328010559082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,2,balanced,0.04520533482233683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,4,power_law_1.2,0.035750401020050046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.2,0.08369920253753663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,2,balanced,0.0446720023949941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,4,power_law_1.2,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.2,0.1021888017654419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,2,balanced,0.044735997915267944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.2,0.041068801283836366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.2,0.13027199506759643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,2,balanced,0.043738668163617454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.2,0.05429760217666626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.2,0.1388800024986267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,2,balanced,0.04494933287302653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.2,0.05989120006561279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.2,0.2612607955932617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,2,balanced,0.04494933287302653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.2,0.07945600152015686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.2,0.2974208116531372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,2,balanced,0.04515733321507772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.2,0.08835840225219727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.2,0.4170368194580078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,2,balanced,0.04451199869314829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.2,0.12485120296478272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.2,0.4874879837036133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,2,balanced,0.049360002080599465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.2,0.15054080486297608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,4,power_law_1.2,0.6442624092102051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,2,balanced,0.050474668542544045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.2,0.21235198974609376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,4,power_law_1.2,1.1847999572753907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,2,balanced,0.04909333089987437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.2,0.2732736110687256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,4,power_law_1.2,2.3395456314086913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,2,balanced,0.05704533557097117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,4,power_law_1.2,0.334003210067749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,2,balanced,0.05973333120346069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,4,power_law_1.2,0.5197184085845947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,2,balanced,0.0714026689529419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,4,power_law_1.2,1.0057151794433594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,2,balanced,0.08090666433175404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,2,balanced,0.10206400354703267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,2,balanced,0.11866666873296101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,2,balanced,0.15851733088493347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,2,balanced,0.18982932964960733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,2,balanced,0.036015999813874565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,2,balanced,0.2672320008277893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,2,balanced,0.03722666700681051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,2,balanced,0.039503999054431915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,2,balanced,0.04048533240954081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,2,balanced,0.04274666806062063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,2,balanced,0.043322667479515076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,2,balanced,0.0409706657131513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,2,balanced,0.33535468578338623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,2,balanced,0.04228266576925913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,2,balanced,0.04309333364168803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,2,balanced,0.043098668257395424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,2,balanced,0.04293866455554962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,2,balanced,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,2,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,2,balanced,0.41278934478759766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,2,balanced,0.04294399917125702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,2,balanced,0.04874666531880697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,2,balanced,0.05086400111516317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,2,balanced,0.051088000337282814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,2,balanced,0.0573226660490036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,2,balanced,0.06084266801675161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,2,balanced,0.07483200232187907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,2,balanced,0.6259040037790934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,2,balanced,0.0919040044148763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,2,balanced,0.12498666842778523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,2,balanced,0.14474133650461832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,2,balanced,0.19988266626993814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,2,balanced,0.2476266622543335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,2,balanced,0.3563093344370524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,2,balanced,1.2057279745737712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,2,balanced,0.4573119878768921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,2,balanced,0.5633279879887899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,2,power_law_1.01,0.038515201210975646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,2,balanced,0.8775786558787028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,2,power_law_1.01,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,2,power_law_1.01,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,2,power_law_1.01,0.044582399725914004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,2,power_law_1.01,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,2,power_law_1.01,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,2,power_law_1.01,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,2,power_law_1.01,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,2,power_law_1.01,0.042905598878860474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,2,power_law_1.01,0.044121599197387694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,2,power_law_1.01,0.036556801199913024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,2,power_law_1.01,0.044684800505638125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,2,power_law_1.01,0.03716480135917664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,2,power_law_1.01,0.04583039879798889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,2,power_law_1.01,0.037651199102401736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,2,power_law_1.01,0.04572800099849701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,2,power_law_1.01,0.040064001083374025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,2,power_law_1.01,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,2,balanced,1.7089120546976726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,2,power_law_1.01,0.04056319892406464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,2,power_law_1.01,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,2,power_law_1.01,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,2,power_law_1.01,0.05185920000076294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,2,power_law_1.01,0.040659201145172116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,2,power_law_1.01,0.05663359761238098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,2,power_law_1.01,0.04200319945812225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,2,power_law_1.01,0.06451200246810913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,2,power_law_1.01,0.04207360148429871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,2,power_law_1.01,0.07260159850120544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,2,power_law_1.01,0.04188160002231598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,2,power_law_1.01,0.09107199907302857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,2,power_law_1.01,0.10261759757995606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,2,power_law_1.01,0.04142720103263855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,2,power_law_1.01,0.04273920059204102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,2,power_law_1.01,0.11914880275726318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,2,power_law_1.01,0.04279040098190308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,2,power_law_1.01,0.045875200629234315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,2,power_law_1.01,0.049542400240898135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,2,power_law_1.01,0.05278720259666443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,2,power_law_1.01,0.054067200422286986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,2,power_law_1.01,0.06060799956321716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,2,power_law_1.01,0.0698303997516632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,2,power_law_1.01,0.08711040019989014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,2,power_law_1.01,0.10924160480499268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,2,power_law_1.01,0.1415295958518982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,2,power_law_1.01,0.16876800060272218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,2,power_law_1.01,0.23984639644622802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,2,power_law_1.01,0.32727038860321045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,2,power_law_1.01,0.4884607791900635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,2,power_law_1.01,0.5947455883026123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,2,power_law_1.01,0.7522240161895752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,2,power_law_1.01,1.269536018371582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,2,power_law_1.01,2.2596672058105467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,2,power_law_1.2,0.038624000549316403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,2,power_law_1.2,0.040275201201438904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,2,power_law_1.2,0.040556800365447995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,2,power_law_1.2,0.044761601090431216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,2,power_law_1.2,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,2,power_law_1.2,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,2,power_law_1.2,0.044838398694992065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,2,power_law_1.2,0.04360319972038269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,2,power_law_1.2,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,2,power_law_1.2,0.04414080083370209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,2,power_law_1.2,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,2,power_law_1.2,0.04581120014190674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,2,power_law_1.2,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,2,power_law_1.2,0.045721599459648134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,2,power_law_1.2,0.05118719935417175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,2,power_law_1.2,0.05258240103721619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,2,power_law_1.2,0.057145601511001586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,2,power_law_1.2,0.06452479958534241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,2,power_law_1.2,0.07167360186576843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,2,power_law_1.2,0.09060479998588562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,2,power_law_1.2,0.10716160535812377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,2,power_law_1.2,0.13684480190277098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,2,power_law_1.2,0.15994240045547486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,2,power_law_1.2,0.20432000160217284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,2,power_law_1.2,0.29528961181640623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,2,power_law_1.2,0.37450881004333497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,2,power_law_1.2,0.4647039890289307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,2,power_law_1.2,0.6492479801177978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,2,power_law_1.2,0.9417856216430665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,2,power_law_1.2,1.6729215621948241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,2,power_law_1.2,0.03529599905014038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,2,power_law_1.2,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,2,power_law_1.2,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,2,power_law_1.2,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,2,power_law_1.2,0.040755200386047366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,2,power_law_1.2,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,2,power_law_1.2,0.04097920060157776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,2,power_law_1.2,0.042534399032592776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,2,power_law_1.2,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,2,power_law_1.2,0.041407999396324155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,2,power_law_1.2,0.04296959936618805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,2,power_law_1.2,0.04336000084877014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,2,power_law_1.2,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,2,power_law_1.2,0.04467839896678925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,2,power_law_1.2,0.05082240104675293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,2,power_law_1.2,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,2,power_law_1.2,0.05374720096588135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,2,power_law_1.2,0.06476160287857055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,2,power_law_1.2,0.0662015974521637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,2,power_law_1.2,0.0865343987941742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,2,power_law_1.2,0.11465599536895751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,2,power_law_1.2,0.15123840570449829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,2,power_law_1.2,0.16391680240631104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,2,power_law_1.2,0.2569983959197998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,2,power_law_1.2,0.3070784091949463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,2,power_law_1.2,0.45751042366027833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,8,balanced,0.06915733218193054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,2,power_law_1.2,0.5895552158355712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,8,balanced,0.06729599833488464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,2,power_law_1.2,0.782092809677124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,8,balanced,0.06403199831644694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,2,power_law_1.2,1.1903936386108398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,8,balanced,0.06923733154932658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,8,balanced,0.06758399804433186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,2,power_law_1.2,2.2508352279663084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,8,balanced,0.06664533416430156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,8,balanced,0.06573866804440816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,8,balanced,0.065610667069753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,8,balanced,0.06700266897678375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,8,balanced,0.06404266754786174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,8,balanced,0.06611733138561249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,8,balanced,0.06198399762312571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,8,balanced,0.060346667965253196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,8,balanced,0.058693334460258484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,8,balanced,0.0603359987338384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,8,balanced,0.059802666306495667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,8,balanced,0.062352001667022705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,8,balanced,0.06868266562620799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,8,balanced,0.07171200215816498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,8,balanced,0.07798933486143748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,8,balanced,0.08927466471989949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,8,balanced,0.11634666721026103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,8,balanced,0.12761066357294717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,2,power_law_1.01,0.14440319538116456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,2,power_law_1.01,0.19587199687957763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,8,balanced,0.16808533668518066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,2,power_law_1.01,0.27891199588775634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,2,power_law_1.01,0.34214398860931394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,2,power_law_1.01,0.4757376194000244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,8,balanced,0.21093867222468057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,2,power_law_1.01,0.6265408039093018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,2,power_law_1.01,0.8257216453552246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,2,power_law_1.01,1.6890560150146485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,8,balanced,0.2924586733182271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,8,balanced,0.05026133358478546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,8,balanced,0.04925866425037384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,8,balanced,0.3625013430913289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,8,balanced,0.0513866643110911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,8,balanced,0.050800000627835594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,8,balanced,0.05283733208974203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,8,balanced,0.05082666873931885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,8,balanced,0.050554667909940086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,8,balanced,0.051967998345692955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,8,balanced,0.04994133114814758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,8,balanced,0.4402773380279541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,8,balanced,0.05102399984995524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,8,balanced,0.05049600203831991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,8,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,8,balanced,0.050757333636283875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,8,balanced,0.048810665806134544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,8,balanced,0.050698667764663696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,8,balanced,0.053642665346463524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,8,balanced,0.05461333195368449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,8,balanced,0.06113600234190623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,8,balanced,0.668890635172526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,8,balanced,0.0629066675901413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,8,balanced,0.07964799801508586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,8,balanced,0.09797333677609761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,8,balanced,0.1290880044301351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,8,balanced,0.15324800213178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,8,balanced,0.2060319979985555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,8,balanced,0.2595679958661397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,8,balanced,0.364896019299825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,8,balanced,1.2844266891479492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,8,balanced,0.46344534556070965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,8,balanced,0.04841599861780802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,8,balanced,0.04894933104515076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,8,balanced,0.5674933195114136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,8,balanced,0.047269334395726524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,8,balanced,0.051669334371884666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,8,balanced,0.05093333125114441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,8,balanced,0.048538664976755776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,8,balanced,0.05303466816743215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,8,balanced,0.049322664737701416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,8,balanced,0.8782506783803304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,8,balanced,0.05019199848175049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,8,balanced,0.04952533543109894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,8,balanced,0.050288001696268715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,8,balanced,0.0517546683549881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,8,balanced,0.046762665112813316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,8,balanced,0.04853333532810211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,8,balanced,0.050794666012128196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,8,balanced,0.04983466863632202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,8,balanced,0.05087466537952423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,8,balanced,0.05985599756240845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,8,balanced,0.06098666787147522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,8,balanced,0.06790400048096974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,8,balanced,0.07723199824492137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,8,balanced,1.704410711924235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,8,balanced,0.10002133250236511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,8,balanced,0.11983999609947205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,8,balanced,0.15124799807866415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,8,balanced,0.1823199987411499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,8,balanced,0.02120000123977661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,8,balanced,0.021503999829292297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,8,balanced,0.022613334159056347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,8,balanced,0.2533386747042338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,8,balanced,0.024218666056791942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,8,balanced,0.022677332162857056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,8,balanced,0.024432001014550526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,8,balanced,0.04827199876308441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,8,balanced,0.048122664292653404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,8,balanced,0.32098132371902466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,8,balanced,0.038949333131313324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,8,balanced,0.036703998843828835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,8,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,8,balanced,0.04365866879622141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,8,balanced,0.04089066634575526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,8,balanced,0.38706668217976886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,8,balanced,0.03897066662708918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,8,balanced,0.04505600035190582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,8,balanced,0.04040000090996424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,8,balanced,0.044138665000597634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,8,balanced,0.057071998715400696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,8,balanced,0.061008001367251076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,8,balanced,0.5885173479715983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,8,balanced,0.08145066599051158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,8,balanced,0.10073066751162212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,8,balanced,0.1344373325506846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,8,balanced,0.16024000446001688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,8,balanced,0.2314079999923706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,8,balanced,0.29148266712824505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,8,balanced,1.1299680074055989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,8,balanced,0.4201226631800334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,8,power_law_1.01,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,8,power_law_1.01,0.06346240043640136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,8,power_law_1.01,0.0632960021495819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,8,power_law_1.01,0.06748800277709961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,8,balanced,0.5504266818364462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,8,power_law_1.01,0.06345599889755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,8,power_law_1.01,0.0623744010925293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,8,power_law_1.01,0.05882880091667175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,8,power_law_1.01,0.057657599449157715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,8,power_law_1.01,0.05568000078201294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,8,balanced,0.6808373133341471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,8,power_law_1.01,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,8,power_law_1.01,0.05185920000076294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,8,power_law_1.01,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,8,power_law_1.01,0.06227840185165405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,8,power_law_1.01,0.06216319799423218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,8,power_law_1.01,0.06544640064239501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,8,balanced,1.0664373238881428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,8,power_law_1.01,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,8,power_law_1.01,0.07738239765167236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,8,power_law_1.01,0.0908352017402649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,8,power_law_1.01,0.1045632004737854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,8,power_law_1.01,0.13573119640350342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,8,power_law_1.01,0.1804927945137024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,8,power_law_1.01,0.2307584047317505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,8,power_law_1.01,0.2784895896911621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,8,power_law_1.01,0.404256010055542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,8,power_law_1.01,0.5607999801635742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,8,power_law_1.01,0.7652607917785644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,8,power_law_1.01,1.013094425201416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,8,balanced,2.0941173235575357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,8,power_law_1.01,1.3219072341918945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,8,power_law_1.01,1.9614143371582031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,8,power_law_1.01,4.216025543212891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,8,power_law_1.01,0.047654399275779726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,8,power_law_1.01,0.04789760112762451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,8,power_law_1.01,0.04878720045089722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,8,power_law_1.01,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,8,power_law_1.01,0.049958398938179015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,8,power_law_1.01,0.0500544011592865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,8,power_law_1.01,0.04865919947624207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,8,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,8,power_law_1.01,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,8,power_law_1.01,0.04645119905471802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,8,power_law_1.01,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,8,power_law_1.01,0.0490880012512207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,8,power_law_1.01,0.04916479885578155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,8,power_law_1.01,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,8,power_law_1.01,0.058432000875473025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,8,power_law_1.01,0.059059202671051025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,8,power_law_1.01,0.06696320176124573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,8,power_law_1.01,0.08131840229034423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,8,power_law_1.01,0.09362559914588928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,8,power_law_1.01,0.11727360486984253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,8,power_law_1.01,0.14932479858398437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,8,power_law_1.01,0.19748480319976808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,8,power_law_1.01,0.25876479148864745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,8,power_law_1.01,0.3588671922683716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,8,power_law_1.01,0.4577216148376465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,8,power_law_1.01,0.626201581954956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,8,power_law_1.01,0.8816127777099609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,8,power_law_1.01,1.0634495735168457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,8,power_law_1.01,1.764748764038086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,8,power_law_1.01,3.508051300048828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,8,power_law_1.01,0.047539201378822324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,8,power_law_1.01,0.04743039906024933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,8,power_law_1.01,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,8,power_law_1.01,0.048793599009513855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,8,power_law_1.01,0.047219198942184445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,8,power_law_1.01,0.04780159890651703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,8,power_law_1.01,0.046982398629188536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,8,power_law_1.01,0.046060800552368164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,8,power_law_1.01,0.04389120042324066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,8,power_law_1.01,0.043584001064300534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,8,power_law_1.01,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,8,power_law_1.01,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,8,power_law_1.01,0.047225600481033324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,8,power_law_1.01,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,8,power_law_1.01,0.056492799520492555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,8,power_law_1.01,0.05687040090560913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,8,power_law_1.01,0.06291840076446534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,8,power_law_1.01,0.06822400093078614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.01,0.0832256019115448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.01,0.10428800582885742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.01,0.12667520046234132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.01,0.16753920316696166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.01,0.2079616069793701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.01,0.2690623998641968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.01,0.3579456090927124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.01,0.5315328121185303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.01,0.6648575782775878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,8,power_law_1.01,0.01971839964389801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,8,power_law_1.01,0.8565312385559082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,8,power_law_1.01,0.02195200026035309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,8,power_law_1.01,1.3064448356628418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,8,power_law_1.01,0.021190400421619415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,8,power_law_1.01,0.02136320024728775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,8,power_law_1.01,2.5640447616577147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,8,power_law_1.01,0.02080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,8,power_law_1.01,0.023871999979019166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,8,power_law_1.01,0.04699519872665405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,8,power_law_1.01,0.04737280011177063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,8,power_law_1.01,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,8,power_law_1.01,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,8,power_law_1.01,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,8,power_law_1.01,0.04350079894065857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,8,power_law_1.01,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,8,power_law_1.01,0.03710080087184906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,8,power_law_1.01,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,8,power_law_1.01,0.039366400241851805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,8,power_law_1.01,0.04309119880199432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,8,power_law_1.01,0.05571200251579285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.01,0.06047999858856201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.01,0.07993599772453308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.01,0.10116480588912964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.01,0.13383040428161622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.01,0.16089600324630737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.01,0.23063039779663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.01,0.2918463945388794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.01,0.42165122032165525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.01,0.5482624053955079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,8,power_law_1.01,0.6789760112762451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,8,power_law_1.01,1.0637120246887206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,8,power_law_1.01,2.0840192794799806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,8,power_law_1.2,0.047366398572921756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,8,power_law_1.2,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,8,power_law_1.2,0.048467200994491574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,8,power_law_1.2,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,8,power_law_1.2,0.05015040040016174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,8,power_law_1.2,0.0493120014667511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,8,power_law_1.2,0.04745599925518036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,8,power_law_1.2,0.047276800870895384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,8,power_law_1.2,0.045459198951721194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,8,power_law_1.2,0.045177599787712096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,8,power_law_1.2,0.04568960070610047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,8,power_law_1.2,0.04800640046596527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,8,power_law_1.2,0.0482367992401123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,8,power_law_1.2,0.04967679977416992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,8,power_law_1.2,0.05486720204353333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,8,power_law_1.2,0.05730559825897217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,8,power_law_1.2,0.06471679806709289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,8,power_law_1.2,0.0828544020652771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,8,power_law_1.2,0.09368320107460022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,8,power_law_1.2,0.12208640575408936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,8,power_law_1.2,0.14826879501342774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,8,power_law_1.2,0.20167040824890137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,8,power_law_1.2,0.2562880039215088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,8,power_law_1.2,0.3572416067123413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,8,power_law_1.2,0.4745664119720459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,8,power_law_1.2,0.6603263854980469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,8,power_law_1.2,0.9007231712341308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,8,power_law_1.2,1.1239999771118163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,8,power_law_1.2,1.7714752197265624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,8,power_law_1.2,3.381024169921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,8,power_law_1.2,0.04689919948577881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,8,power_law_1.2,0.04750719964504242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,8,power_law_1.2,0.046726399660110475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,8,power_law_1.2,0.05008640289306641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,8,power_law_1.2,0.04876160025596619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,8,power_law_1.2,0.04845440089702606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,8,power_law_1.2,0.0470335990190506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,8,power_law_1.2,0.04583039879798889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,8,power_law_1.2,0.04320639967918396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,8,power_law_1.2,0.044870400428771974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,8,power_law_1.2,0.04167680144309997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,8,power_law_1.2,0.05203840136528015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,8,power_law_1.2,0.04860160052776337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,8,power_law_1.2,0.04595839977264404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,8,power_law_1.2,0.058713597059249875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,8,power_law_1.2,0.056966400146484374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,8,power_law_1.2,0.06056320071220398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,8,power_law_1.2,0.06818559765815735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.2,0.08470399975776673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.2,0.10337920188903808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.2,0.1186560034751892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.2,0.1688256025314331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.2,0.20263679027557374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.2,0.29123198986053467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.2,0.37342081069946287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.2,0.537497615814209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.2,0.670143985748291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,8,power_law_1.2,0.8662528038024903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,8,power_law_1.2,1.381875228881836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,8,power_law_1.2,2.6619903564453127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,8,power_law_1.2,0.02104319930076599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,8,power_law_1.2,0.021561600267887115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,8,power_law_1.2,0.021638399362564086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,8,power_law_1.2,0.021503999829292297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,8,power_law_1.2,0.023052799701690673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,8,power_law_1.2,0.023846399784088135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,8,power_law_1.2,0.04705919921398163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,8,power_law_1.2,0.04794879853725433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,8,power_law_1.2,0.037887999415397645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,8,power_law_1.2,0.035334399342536925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,8,power_law_1.2,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,8,power_law_1.2,0.04128639996051788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,8,power_law_1.2,0.0420415997505188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,8,power_law_1.2,0.061267197132110596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,8,power_law_1.2,0.036550399661064145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,8,power_law_1.2,0.06295040249824524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,8,power_law_1.2,0.044223999977111815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,8,power_law_1.2,0.06230400204658508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,8,power_law_1.2,0.039443200826644896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,8,power_law_1.2,0.06595199704170226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,8,power_law_1.2,0.04384639859199524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,8,power_law_1.2,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,8,power_law_1.2,0.05752320289611816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,8,power_law_1.2,0.0624064028263092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.2,0.06053119897842407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,8,power_law_1.2,0.060499197244644164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.2,0.0808896005153656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,8,power_law_1.2,0.057792001962661745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.2,0.10010240077972413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,8,power_law_1.2,0.0541055977344513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.2,0.134662401676178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,8,power_law_1.2,0.05262079834938049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.2,0.1597759962081909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,8,power_law_1.2,0.052249598503112796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.2,0.23055999279022216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,8,power_law_1.2,0.06381440162658691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.2,0.29063680171966555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,8,power_law_1.2,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.2,0.4214015960693359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,8,power_law_1.2,0.059724801778793336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.2,0.5507584095001221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,8,power_law_1.2,0.07264639735221863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,8,power_law_1.2,0.6797696113586426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,8,power_law_1.2,0.07045120000839233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,8,power_law_1.2,1.0653504371643066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,8,power_law_1.2,0.07783039808273315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,4,balanced,0.05107733110586802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,8,power_law_1.2,2.0914688110351562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,8,power_law_1.2,0.09781119823455811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,4,balanced,0.0517439991235733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,8,power_law_1.2,0.1090880036354065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,4,balanced,0.05852266649405161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,8,power_law_1.2,0.14332159757614135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,4,balanced,0.05937066674232483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,8,power_law_1.2,0.17640960216522217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,4,balanced,0.057706668972969055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,8,power_law_1.2,0.23907198905944824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,4,balanced,0.06347733239332835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,8,power_law_1.2,0.299020791053772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,4,balanced,0.05640000104904175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,4,balanced,0.05425066749254862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,8,power_law_1.2,0.4389887809753418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,4,balanced,0.05715199808279673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,8,power_law_1.2,0.5411392211914062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,4,balanced,0.05994666616121928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,8,power_law_1.2,0.7801023960113526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,4,balanced,0.05699733396371206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,8,power_law_1.2,1.0689536094665528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,4,balanced,0.056602666775385536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,8,power_law_1.2,1.3286656379699706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,4,balanced,0.05508266886075338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,8,power_law_1.2,1.9479616165161133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,4,balanced,0.055455997586250305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,4,balanced,0.06238399942715963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,8,power_law_1.2,4.2482177734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,4,balanced,0.06079466640949249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,4,balanced,0.06303999821345012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,4,balanced,0.07162133355935414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,4,balanced,0.07650133470694225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,4,balanced,0.0831413318713506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,4,balanced,0.09141332904497783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,4,balanced,0.12055466572443645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,4,balanced,0.13470932841300964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,4,balanced,0.17782400051752725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,4,balanced,0.22025599082310995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,4,balanced,0.3095146616299947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,4,balanced,0.38412801424662274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,4,balanced,0.04106666644414266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,4,balanced,0.04256533086299896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,4,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,4,balanced,0.049173335234324135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,4,balanced,0.47707732518513996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,4,balanced,0.048197334011395775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,4,balanced,0.04922133187452952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,4,balanced,0.05083199838797251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,4,balanced,0.048986668388048805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,4,balanced,0.04921066761016846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,4,balanced,0.049770668148994446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,4,balanced,0.04690133531888326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,4,balanced,0.721898635228475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,4,balanced,0.04709866642951965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,4,balanced,0.05061866839726766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,4,balanced,0.04916800061861674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,4,balanced,0.05271466573079427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,4,balanced,0.05268799761931101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,4,balanced,0.053183997670809426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,4,balanced,0.06515199939409892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,4,balanced,0.06929600238800049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,4,balanced,0.08566932876904805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,4,balanced,0.10355200370152791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,4,balanced,1.406559944152832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,4,balanced,0.13741866747538248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,4,balanced,0.1625333329041799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,4,balanced,0.04145599901676178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,4,balanced,0.22292800744374594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,4,balanced,0.043706665436426796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,4,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,4,balanced,0.049253334601720176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,4,balanced,0.04666133224964142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,4,balanced,0.28037333488464355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,4,balanced,0.04764266808827718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,4,balanced,0.04923733572165171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,4,balanced,0.05153599878152212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,4,balanced,0.04706133405367533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,4,balanced,0.051039998730023704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,4,balanced,0.3981706698735555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,4,balanced,0.048197334011395775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,4,balanced,0.05052266518274943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,4,balanced,0.04894933104515076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,4,balanced,0.04701866706212362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,4,balanced,0.05276266733805338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,4,balanced,0.5093013445536295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,4,balanced,0.050186668833096824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,4,balanced,0.05302399893601736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,4,balanced,0.01966399947802226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,4,balanced,0.02146666745344798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,4,balanced,0.060778667529424034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,4,balanced,0.022672000030676525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,4,balanced,0.06504533191521962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,4,balanced,0.022463999688625336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,4,balanced,0.07386666536331177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,4,balanced,0.022319999833901722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,4,balanced,0.022869333624839783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,4,balanced,0.08763200044631958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,4,balanced,0.6292266845703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,4,balanced,0.025557334224383037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,4,balanced,0.10717866818110149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,4,balanced,0.026309333741664886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,4,balanced,0.04906666775544485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,4,balanced,0.04885333279768626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,4,balanced,0.13008532921473184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,4,balanced,0.04929600159327189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,4,balanced,0.03666666646798452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,4,balanced,0.16859199603398642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,4,balanced,0.03892799963553747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,4,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,4,balanced,0.04027199993530909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,4,balanced,0.2046026587486267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,4,balanced,0.03900266687075297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,4,balanced,0.9733386834462484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,4,balanced,0.04247466723124186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,4,balanced,0.046522667010625206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,4,balanced,0.2870933413505554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,4,balanced,0.04868266483147939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,4,balanced,0.06749866902828217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,4,balanced,0.07529599964618683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,4,balanced,0.09954133629798889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,4,balanced,0.3616746664047241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,4,balanced,0.12292266885439555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,4,balanced,0.16965866088867188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,4,balanced,0.4422293504079183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,4,balanced,0.204912006855011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,4,balanced,1.8946293195088704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,4,balanced,0.2961973349253337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,4,balanced,0.6773280302683512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,4,balanced,0.3815679947535197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,4,balanced,0.4647200107574463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,4,balanced,1.3085280259450276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,4,balanced,0.7255573272705078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,4,power_law_1.01,0.055852800607681274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,4,power_law_1.01,0.05398399829864502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,4,power_law_1.01,0.05455999970436096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,4,power_law_1.01,0.05724160075187683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,4,power_law_1.01,0.0571008026599884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,4,power_law_1.01,0.05735679864883423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,4,power_law_1.01,0.055827200412750244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,4,power_law_1.01,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,4,power_law_1.01,0.05601919889450073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,4,power_law_1.01,0.05486080050468445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,4,balanced,1.4209280014038086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,4,power_law_1.01,0.05598719716072083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,4,power_law_1.01,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,4,power_law_1.01,0.058905601501464844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,4,power_law_1.01,0.06260480284690857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,4,power_law_1.01,0.06991999745368957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,4,power_law_1.01,0.0720192015171051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,4,power_law_1.01,0.07452800273895263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,4,power_law_1.01,0.0869376003742218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,4,power_law_1.01,0.09173120260238647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,4,power_law_1.01,0.12499840259552002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,4,power_law_1.01,0.16120320558547974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,4,power_law_1.01,0.18071039915084838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,4,power_law_1.01,0.21619839668273927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,4,power_law_1.01,0.2924992084503174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,4,power_law_1.01,0.40524802207946775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,4,power_law_1.01,0.6760128021240235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,4,power_law_1.01,0.9496383666992188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,4,power_law_1.01,0.998265552520752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,4,power_law_1.01,1.54203519821167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,4,power_law_1.01,3.1732608795166017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,4,power_law_1.01,0.04206719994544983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,4,power_law_1.01,0.04402559995651245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,4,power_law_1.01,0.046214398741722104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,4,power_law_1.01,0.045286399126052854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,4,power_law_1.01,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,4,power_law_1.01,0.04672000110149384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,4,power_law_1.01,0.047270399332046506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,4,power_law_1.01,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,4,power_law_1.01,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,4,power_law_1.01,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,4,power_law_1.01,0.04754559993743897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,4,power_law_1.01,0.04757120013237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,4,power_law_1.01,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,4,power_law_1.01,0.04904319941997528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,4,power_law_1.01,0.04549759924411774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,4,power_law_1.01,0.050937598943710326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,4,power_law_1.01,0.046009600162506104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,4,power_law_1.01,0.05694720149040222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,4,power_law_1.01,0.04664320051670075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,4,power_law_1.01,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,4,power_law_1.01,0.04799999892711639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,4,power_law_1.01,0.06202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,4,power_law_1.01,0.048870399594306946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,4,power_law_1.01,0.019468800723552705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,4,power_law_1.01,0.07493119835853576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,4,power_law_1.01,0.04854399859905243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,4,power_law_1.01,0.08884479999542236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,4,power_law_1.01,0.019942399859428406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,4,power_law_1.01,0.04838399887084961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,4,power_law_1.01,0.020448000729084016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,4,power_law_1.01,0.12253439426422119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,4,power_law_1.01,0.04764800071716309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,4,power_law_1.01,0.020179200172424316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,4,power_law_1.01,0.04755200147628784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,4,power_law_1.01,0.13036160469055175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,4,power_law_1.01,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,4,power_law_1.01,0.02019200026988983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,4,power_law_1.01,0.1859392046928406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,4,power_law_1.01,0.04855040013790131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,4,power_law_1.01,0.020851199328899384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,4,power_law_1.01,0.2420351982116699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,4,power_law_1.01,0.048665601015090945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,4,power_law_1.01,0.023161600530147552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,4,power_law_1.01,0.3474816083908081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,4,power_law_1.01,0.05077120065689087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,4,power_law_1.01,0.023366400599479677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,4,power_law_1.01,0.4442431926727295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,4,power_law_1.01,0.05583360195159912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,4,power_law_1.01,0.04754559993743897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,4,power_law_1.01,0.5919680118560791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,4,power_law_1.01,0.05982080101966858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,4,power_law_1.01,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,4,power_law_1.01,0.8137855529785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,4,power_law_1.01,0.06230400204658508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,4,power_law_1.01,0.047891199588775635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,4,power_law_1.01,0.8768256187438965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,4,power_law_1.01,0.06981760263442993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,4,power_law_1.01,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,4,power_law_1.01,1.7512832641601563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.01,0.08885120153427124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,4,power_law_1.01,0.03660799860954285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,4,power_law_1.01,3.294790267944336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.01,0.10424319505691529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,4,power_law_1.01,0.03713920116424561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.01,0.11696000099182129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,4,power_law_1.01,0.03930239975452423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.01,0.17069439888000487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,4,power_law_1.01,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.01,0.19925119876861572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,4,power_law_1.01,0.040249601006507874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.01,0.30404479503631593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,4,power_law_1.01,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.01,0.04656639993190766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.01,0.06536960005760192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.01,0.07533439993858337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.01,0.09882879853248597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.01,0.12359039783477783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.01,0.16931840181350707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,4,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.01,0.2056960105895996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,4,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.01,0.29838080406188966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,4,power_law_1.2,0.05882880091667175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.01,0.3804991960525513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,4,power_law_1.2,0.05924479961395264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,4,power_law_1.01,0.46370558738708495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,4,power_law_1.2,0.05653759837150574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,4,power_law_1.2,0.05440000295639038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,4,power_law_1.01,0.7254208087921142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,4,power_law_1.2,0.055878400802612305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,4,power_law_1.01,1.4208703994750977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,4,power_law_1.2,0.056358402967453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,4,power_law_1.2,0.05669119954109192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,4,power_law_1.2,0.05730559825897217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,4,power_law_1.2,0.055270397663116456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,4,power_law_1.2,0.06212480068206787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,4,power_law_1.2,0.06311039924621582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,4,power_law_1.2,0.063372802734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,4,power_law_1.2,0.06863999962806702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,4,power_law_1.2,0.07257599830627441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,4,power_law_1.2,0.07758079767227173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,4,power_law_1.2,0.08906880021095276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,4,power_law_1.2,0.10551040172576905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,4,power_law_1.2,0.13523839712142943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,4,power_law_1.2,0.15675519704818724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,4,power_law_1.2,0.17871999740600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,4,power_law_1.2,0.26572160720825194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,4,power_law_1.2,0.3897599935531616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,4,power_law_1.2,0.44155521392822267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,4,power_law_1.2,0.5982143878936768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,4,power_law_1.2,1.0754176139831544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,4,power_law_1.2,1.3726783752441407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,4,power_law_1.2,1.835385513305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,4,power_law_1.2,3.7410240173339844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,4,power_law_1.2,0.042982399463653564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,4,power_law_1.2,0.04222719967365265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,4,power_law_1.2,0.043993601202964784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,4,power_law_1.2,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,4,power_law_1.2,0.04611839950084686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,4,power_law_1.2,0.04745599925518036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,4,power_law_1.2,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,4,power_law_1.2,0.04659200012683869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,4,power_law_1.2,0.04528000056743622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,4,power_law_1.2,0.047513601183891294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,4,power_law_1.2,0.047679999470710756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,4,power_law_1.2,0.04688639938831329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,4,power_law_1.2,0.04280959963798523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,4,power_law_1.2,0.04867840111255646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,4,power_law_1.2,0.04294399917125702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,4,power_law_1.2,0.05106559991836548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,4,power_law_1.2,0.044268798828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,4,power_law_1.2,0.05852159857749939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,4,power_law_1.2,0.049728000164031984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,4,power_law_1.2,0.05939840078353882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,4,power_law_1.2,0.048153600096702574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,4,power_law_1.2,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,4,power_law_1.2,0.04595839977264404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,4,power_law_1.2,0.08024320006370544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,4,power_law_1.2,0.04561919867992401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,4,power_law_1.2,0.0887615978717804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,4,power_law_1.2,0.04755840003490448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,4,power_law_1.2,0.11843839883804322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,4,power_law_1.2,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,4,power_law_1.2,0.15036799907684326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,4,power_law_1.2,0.04624640047550201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,4,power_law_1.2,0.20008320808410646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,4,power_law_1.2,0.04721280038356781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,4,power_law_1.2,0.2641279935836792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,4,power_law_1.2,0.048876801133155824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,4,power_law_1.2,0.3991039991378784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,4,power_law_1.2,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,4,power_law_1.2,0.43160319328308105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,4,power_law_1.2,0.050995200872421265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,4,power_law_1.2,0.6577727794647217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,4,power_law_1.2,0.058899199962615965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,4,power_law_1.2,0.8845312118530273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,4,power_law_1.2,0.055936002731323244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,4,power_law_1.2,1.168825626373291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,4,power_law_1.2,0.06349440217018128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,4,power_law_1.2,1.7997695922851562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,4,power_law_1.2,0.0730239987373352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,4,power_law_1.2,3.951161575317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.2,0.08570240139961242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.2,0.11361919641494751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.2,0.12419840097427368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.2,0.15191680192947388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.2,0.19811840057373048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.2,0.27984640598297117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.2,0.3618367910385132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.2,0.4747007846832275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.2,0.7999104022979736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,4,power_law_1.2,0.9707200050354003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,4,power_law_1.2,0.01963520050048828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,4,power_law_1.2,1.2536447525024415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,4,power_law_1.2,0.021216000616550445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,4,power_law_1.2,3.2061824798583984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,4,power_law_1.2,0.020960000157356263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,4,power_law_1.2,0.02237440049648285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,4,power_law_1.2,0.02111999988555908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,4,power_law_1.2,0.022291199862957002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,4,power_law_1.2,0.02503040134906769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.01,0.38421120643615725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,4,power_law_1.2,0.02563199996948242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.01,0.5106944084167481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,4,power_law_1.2,0.04715520143508911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,4,power_law_1.2,0.047193598747253415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.01,0.6196159839630127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,4,power_law_1.2,0.048019200563430786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,4,power_law_1.01,0.7651711940765381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,4,power_law_1.2,0.036652800440788266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,4,power_law_1.01,1.4313088417053224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,4,power_law_1.2,0.037484800815582274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,4,power_law_1.01,2.4540544509887696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,4,power_law_1.2,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,4,power_law_1.2,0.03858560025691986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,4,power_law_1.2,0.03989759981632233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,4,power_law_1.2,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,4,power_law_1.2,0.04506239891052246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.2,0.04774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.2,0.06833919882774353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.2,0.07364479899406433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.2,0.09797760248184204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.2,0.12264319658279418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.2,0.16859519481658936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.2,0.2046720027923584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.2,0.29706239700317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.2,0.38125441074371336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,4,power_law_1.2,0.46436481475830077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,4,power_law_1.2,0.7241792201995849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,4,power_law_1.2,1.41463041305542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,2,balanced,0.044490665197372437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,2,balanced,0.04740266501903534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,2,balanced,0.05351466437180837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,2,balanced,0.05956799785296122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,2,balanced,0.05394133428732554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,2,balanced,0.05469333132108053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,2,balanced,0.05519466598828634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,2,balanced,0.05483733117580414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,2,balanced,0.05333866675694784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,2,balanced,0.05490666627883911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,2,balanced,0.055013333757718406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,2,balanced,0.055045331517855324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,2,balanced,0.03787733366092046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,2,balanced,0.05705066521962484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,2,balanced,0.04091733445723852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,2,balanced,0.04398400088151296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,2,balanced,0.05721066892147064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,2,balanced,0.0452159990866979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,2,balanced,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,2,balanced,0.04717333118120829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,2,balanced,0.06326933205127716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,2,balanced,0.048112000028292336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,2,balanced,0.0652106652657191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,2,balanced,0.04507199923197428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,2,balanced,0.07702399790287018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,2,balanced,0.04915733138720194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,2,balanced,0.04711999992529551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,2,balanced,0.038431999584039055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,2,balanced,0.07998399933179219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,2,balanced,0.047210668524106346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,2,balanced,0.045461331804593406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,2,balanced,0.09290132919947307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,2,balanced,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,2,balanced,0.04930133124192556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,2,balanced,0.049269333481788635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,2,balanced,0.04448533554871877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,2,balanced,0.102101335922877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,2,balanced,0.04837866624196371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,2,balanced,0.04886400202910105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,2,balanced,0.0528106689453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,2,balanced,0.1347040037314097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,2,balanced,0.04860800007979075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,2,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,2,balanced,0.05318933228651682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,2,balanced,0.15516799688339233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,2,balanced,0.051226665576299034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,2,balanced,0.055306668082873024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,2,balanced,0.05115200082461039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,2,balanced,0.05715199808279673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,2,balanced,0.0488373339176178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,2,balanced,0.0691840002934138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,2,balanced,0.21163199345270792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,2,balanced,0.049226666490236916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,2,balanced,0.07745066781838734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,2,balanced,0.04817600051561991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,2,balanced,0.055071999629338585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,2,balanced,0.10123733679453532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,2,balanced,0.04955733319123586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,2,balanced,0.26597867409388226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,2,balanced,0.11740266283353169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,2,balanced,0.05268266797065735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,2,balanced,0.15391467014948526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,2,balanced,0.05657599866390228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,2,balanced,0.3733919858932495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,2,balanced,0.05693333347638448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,2,balanced,0.18454933166503906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,2,balanced,0.067930668592453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,2,balanced,0.06986133257548015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,2,balanced,0.25945067405700684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,2,balanced,0.08666666348775227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,2,balanced,0.46955732504526776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,2,balanced,0.33027199904123944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,2,balanced,0.09779733419418335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,2,balanced,0.12621866663297018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,2,balanced,0.47577067216237384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,2,balanced,0.5755093495051066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,2,balanced,0.152346670627594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,2,balanced,0.20802666743596396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,2,balanced,0.607040007909139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,2,balanced,0.021317332983016968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,2,balanced,0.02239466706911723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,2,balanced,0.2553013364473979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,2,balanced,0.02231466770172119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,2,balanced,0.8913653691609701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,2,balanced,0.022783999641736347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,2,balanced,0.022522665560245514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,2,balanced,0.7488746643066406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,2,balanced,0.02214933435122172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,2,balanced,0.3671146631240845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,2,balanced,0.022618666291236877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,2,balanced,0.02290133386850357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,2,balanced,0.024661332368850708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,2,balanced,0.024133334557215374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,2,balanced,0.4694826602935791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,2,balanced,0.02651199946800868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,2,balanced,0.04877333343029022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,2,balanced,1.168298641840617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,2,balanced,0.04886400202910105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,2,balanced,0.05114666620890299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,2,balanced,0.038586666186650596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,2,balanced,0.5714720090230306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,2,balanced,0.03846399982770284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,2,balanced,1.7266772588094075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,2,balanced,0.041135999063650765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,2,balanced,0.0447626660267512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,2,balanced,0.04837866624196371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,2,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,2,balanced,0.8903040091196696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,2,balanced,0.06276800235112508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,2,balanced,2.2805333137512207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,2,balanced,0.08593066533406575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,2,balanced,0.10020800431569417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,2,balanced,0.1341813306013743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,2,balanced,0.1679840087890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,2,balanced,1.7216480573018391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,2,balanced,0.23751999934514365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,2,balanced,0.2974453369776408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,2,balanced,0.36794666449228924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,2,balanced,0.5680266618728638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,2,balanced,1.09333332379659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,2,power_law_1.01,0.044588801264762876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,2,power_law_1.01,0.04973439872264862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,2,power_law_1.01,0.05134080052375793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,2,power_law_1.01,0.05356799960136414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,2,power_law_1.01,0.0537280023097992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,2,power_law_1.01,0.052748799324035645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,2,power_law_1.01,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,2,power_law_1.01,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,2,power_law_1.01,0.03709439933300018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,2,power_law_1.01,0.054995197057724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,2,power_law_1.01,0.04099839925765991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,2,power_law_1.01,0.055244797468185426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,2,power_law_1.01,0.04132480025291443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,2,power_law_1.01,0.05418239831924439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,2,power_law_1.01,0.04537599980831146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,2,power_law_1.01,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,2,power_law_1.01,0.058719998598098753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,2,power_law_1.01,0.045388799905776975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,2,power_law_1.01,0.0598143994808197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,2,power_law_1.01,0.04531840085983276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,2,power_law_1.01,0.06726400256156921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,2,power_law_1.01,0.04678399860858917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,2,power_law_1.01,0.0690559983253479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,2,power_law_1.01,0.047091200947761536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,2,power_law_1.01,0.04266240000724793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,2,power_law_1.01,0.07402880191802978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,2,power_law_1.01,0.046777600049972536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,2,power_law_1.01,0.0400191992521286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,2,power_law_1.01,0.07964800000190735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,2,power_law_1.01,0.046463999152183535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,2,power_law_1.01,0.04382080137729645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,2,power_law_1.01,0.08747519850730896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,2,power_law_1.01,0.04645119905471802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,2,power_law_1.01,0.047712001204490664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,2,power_law_1.01,0.11331199407577515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,2,power_law_1.01,0.04789760112762451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,2,power_law_1.01,0.04797439873218536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,2,power_law_1.01,0.049728000164031984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,2,power_law_1.01,0.13093760013580322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,2,power_law_1.01,0.04792959988117218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,2,power_law_1.01,0.049414399266242984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,2,power_law_1.01,0.1747007966041565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,2,power_law_1.01,0.04883840084075928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,2,power_law_1.01,0.05628160238265991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,2,power_law_1.01,0.04819200038909912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,2,power_law_1.01,0.19475200176239013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,2,power_law_1.01,0.058361601829528806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,2,power_law_1.01,0.04793600142002106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,2,power_law_1.01,0.28988161087036135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,2,power_law_1.01,0.04799999892711639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,2,power_law_1.01,0.06312959790229797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,2,power_law_1.01,0.3479871988296509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,2,power_law_1.01,0.04901759922504425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,2,power_law_1.01,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,2,power_law_1.01,0.5104959964752197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,2,power_law_1.01,0.050316798686981204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,2,power_law_1.01,0.09104639887809754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,2,power_law_1.01,0.6912320137023926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,2,power_law_1.01,0.051577597856521606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,2,power_law_1.01,0.11725440025329589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,2,power_law_1.01,0.8247551918029785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,2,power_law_1.01,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,2,power_law_1.01,0.1375167965888977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,2,power_law_1.01,1.3332096099853517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,2,power_law_1.01,0.0588096022605896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,2,power_law_1.01,0.18885120153427123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,2,power_law_1.01,2.5146368026733397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,2,power_law_1.01,0.06047359704971313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,2,power_law_1.01,0.21938560009002686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,2,power_law_1.01,0.06394240260124207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,2,power_law_1.01,0.35433599948883054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,2,power_law_1.01,0.07494400143623352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,2,power_law_1.01,0.40778241157531736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.01,0.08508800268173218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,2,power_law_1.01,0.6131519794464111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.01,0.10006400346755981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,2,power_law_1.01,0.7893311977386475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.01,0.12378239631652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,2,power_law_1.01,0.8895872116088868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.01,0.15794559717178344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,2,power_law_1.01,1.4590975761413574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.01,0.1976639986038208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,2,power_law_1.01,3.068281555175781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.01,0.2921152114868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.01,0.33255040645599365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.01,0.42687358856201174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.01,0.6035647869110108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,2,power_law_1.01,0.6895423889160156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,2,power_law_1.01,1.2444735527038575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,2,power_law_1.01,2.559667205810547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,2,power_law_1.01,0.01916159987449646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,2,power_law_1.01,0.02064639925956726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,2,power_law_1.01,0.020870399475097657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,2,power_law_1.01,0.021216000616550445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,2,power_law_1.01,0.02078080028295517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,2,power_law_1.01,0.020710399746894835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,2,power_law_1.01,0.020319999754428865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,2,power_law_1.01,0.021516799926757812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,2,power_law_1.01,0.023974399268627166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,2,power_law_1.01,0.02446720004081726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,2,power_law_1.01,0.025088000297546386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,2,power_law_1.01,0.047353601455688475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,2,power_law_1.01,0.04839679896831513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,2,power_law_1.01,0.04943360090255737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,2,power_law_1.01,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,2,power_law_1.01,0.03823359906673431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,2,power_law_1.01,0.0403903990983963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,2,power_law_1.01,0.04413439929485321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.01,0.046726399660110475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.01,0.05394560098648071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.01,0.06297600269317627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.01,0.0862335979938507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.01,0.09840000271797181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.01,0.13422080278396606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.01,0.1686527967453003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.01,0.23624320030212403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.01,0.2976576089859009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,2,power_law_1.01,0.3664128065109253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,2,power_law_1.01,0.5682559967041015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,2,power_law_1.01,1.0928640365600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,2,power_law_1.2,0.04707840085029602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,2,power_law_1.2,0.04927360117435455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,2,power_law_1.2,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,2,power_law_1.2,0.054150402545928955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,2,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,2,power_law_1.2,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,2,power_law_1.2,0.05283839702606201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,2,power_law_1.2,0.057145601511001586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,2,power_law_1.2,0.03870719969272614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,2,power_law_1.2,0.05448319911956787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,2,power_law_1.2,0.04167680144309997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,2,power_law_1.2,0.055846399068832396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,2,power_law_1.2,0.04076800048351288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,2,power_law_1.2,0.0549888014793396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,2,power_law_1.2,0.045721599459648134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,2,power_law_1.2,0.05896959900856018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,2,power_law_1.2,0.04599039852619171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,2,power_law_1.2,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,2,power_law_1.2,0.045824000239372255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,2,power_law_1.2,0.06158080101013184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,2,power_law_1.2,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,2,power_law_1.2,0.06736639738082886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,2,power_law_1.2,0.0430400013923645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,2,power_law_1.2,0.0686079978942871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,2,power_law_1.2,0.04151679873466492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,2,power_law_1.2,0.07475200295448303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,2,power_law_1.2,0.048230400681495665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,2,power_law_1.2,0.08296959996223449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,2,power_law_1.2,0.04902400076389313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,2,power_law_1.2,0.09574400186538697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,2,power_law_1.2,0.04945279955863953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,2,power_law_1.2,0.11960959434509277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,2,power_law_1.2,0.04970879852771759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,2,power_law_1.2,0.13225599527359008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,2,power_law_1.2,0.048979198932647704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,2,power_law_1.2,0.18581759929656982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,2,power_law_1.2,0.0495743989944458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,2,power_law_1.2,0.2147200107574463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,2,power_law_1.2,0.049395200610160825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,2,power_law_1.2,0.3172991991043091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,2,power_law_1.2,0.04892799854278564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,2,power_law_1.2,0.3699327945709229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,2,power_law_1.2,0.05241600275039673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,2,power_law_1.2,0.5466623783111573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,2,power_law_1.2,0.05189120173454285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,2,power_law_1.2,0.6882175922393798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,2,power_law_1.2,0.05160319805145264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,2,power_law_1.2,0.9043135643005371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,2,power_law_1.2,0.05804799795150757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,2,power_law_1.2,1.5309696197509766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,2,power_law_1.2,0.060678398609161376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,2,power_law_1.2,2.5238975524902343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,2,power_law_1.2,0.06438400149345398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,2,power_law_1.2,0.07379199862480164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.2,0.08774399757385254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.2,0.10296319723129273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.2,0.12702080011367797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.2,0.16492799520492554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.2,0.19550080299377443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.2,0.2604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.2,0.3611327886581421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.2,0.5371263980865478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.2,0.6475647926330567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,2,power_law_1.2,0.8015872001647949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,2,power_law_1.2,1.3391167640686035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,2,power_law_1.2,2.414758491516113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,2,power_law_1.2,0.019385600090026857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,2,power_law_1.2,0.020524799823760986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,2,power_law_1.2,0.021503999829292297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,2,power_law_1.2,0.02115200012922287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,2,power_law_1.2,0.021171200275421142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,2,power_law_1.2,0.021529600024223328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,2,power_law_1.2,0.02139520049095154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,2,power_law_1.2,0.022355200350284578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,2,power_law_1.2,0.02389120012521744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,2,power_law_1.2,0.02503040134906769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,2,power_law_1.2,0.024563199281692503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,2,power_law_1.2,0.04847359955310822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,2,power_law_1.2,0.04746240079402923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,2,power_law_1.2,0.04925439953804016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,2,power_law_1.2,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,2,power_law_1.2,0.036447998881340024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,2,power_law_1.2,0.04028159976005554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,2,power_law_1.2,0.04268800020217896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.2,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.2,0.05413119792938233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.2,0.06080639958381653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.2,0.08549759984016418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.2,0.09689599871635438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.2,0.13315839767456056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.2,0.16844799518585205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.2,0.2381824016571045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.2,0.2977920055389404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,2,power_law_1.2,0.36740479469299314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,2,power_law_1.2,0.5689407825469971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,2,power_law_1.2,1.0921536445617677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,2,power_law_1.2,0.0457152009010315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,64,balanced,0.08365333080291748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,2,power_law_1.2,0.04673919975757599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,64,balanced,0.08781333764394124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,2,power_law_1.2,0.04684160053730011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,64,balanced,0.0883146623770396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,2,power_law_1.2,0.04633600115776062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,64,balanced,0.08423466483751933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,64,balanced,0.09943999846776326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,2,power_law_1.2,0.04705280065536499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,64,balanced,0.13290133078893027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,2,power_law_1.2,0.04760960042476654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,64,balanced,0.12753066420555115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,2,power_law_1.2,0.048793599009513855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,64,balanced,0.13269866506258646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,2,power_law_1.2,0.05066239833831787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,2,power_law_1.2,0.05740159749984741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,64,balanced,0.13703999916712442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,2,power_law_1.2,0.05902079939842224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,64,balanced,0.13909332950909933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,2,power_law_1.2,0.06518399715423584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,64,balanced,0.13611732920010886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,2,power_law_1.2,0.07888640165328979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,64,balanced,0.13609600067138672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,2,power_law_1.2,0.09000319838523865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,64,balanced,0.13618133465449014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,2,power_law_1.2,0.11153279542922974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,64,balanced,0.13542399803797403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,2,power_law_1.2,0.13262079954147338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,64,balanced,0.06358933448791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,64,balanced,0.14407466848691305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,2,power_law_1.2,0.1859007954597473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,64,balanced,0.06655466556549072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,64,balanced,0.06568000217278798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,2,power_law_1.2,0.24432640075683593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,64,balanced,0.14180266857147217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,64,balanced,0.06752533217271169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,2,power_law_1.2,0.3260672092437744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,64,balanced,0.14301333824793497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,64,balanced,0.07396266857783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,2,power_law_1.2,0.39305601119995115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,64,balanced,0.09688533345858256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,2,power_law_1.2,0.6237696170806885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,64,balanced,0.14972800016403198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,64,balanced,0.0946560005346934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,2,power_law_1.2,0.7819968223571777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,64,balanced,0.09617066383361816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,2,power_law_1.2,0.9944704055786133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,64,balanced,0.1476853291193644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,64,balanced,0.0974079966545105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,2,power_law_1.2,1.4713279724121093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,64,balanced,0.09929600358009338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,2,power_law_1.2,3.1624639511108397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,64,balanced,0.14778666694959006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,64,balanced,0.10621333122253418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,64,balanced,0.10228799780209859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,64,balanced,0.10041600465774536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,64,balanced,0.15903466939926147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,64,balanced,0.10406399766604106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,64,balanced,0.10826133688290913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,64,balanced,0.17857599258422852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,64,balanced,0.11503466963768005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,64,balanced,0.11424533526102702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,64,balanced,0.1234773298104604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,64,balanced,0.19987734158833823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,64,balanced,0.12844799955685934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,64,balanced,0.14287466804186502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,64,balanced,0.060677334666252136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,64,balanced,0.06554133196671803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,64,balanced,0.27164266506830853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,64,balanced,0.16057067116101584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,64,balanced,0.06267733375231425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,64,balanced,0.0636106679836909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,64,balanced,0.07719466586907704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,64,balanced,0.09937600294748943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,64,balanced,0.19658666849136353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,64,balanced,0.09542933106422424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,64,balanced,0.09918933113416036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,64,balanced,0.10002666711807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,64,balanced,0.09390399853388469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,64,balanced,0.09829333424568176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,64,balanced,0.23388799031575522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,64,balanced,0.09821866949399312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,64,balanced,0.0969493289788564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,64,balanced,0.10205333431561787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,64,balanced,0.10387733578681946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,64,balanced,0.10626133282979329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,64,balanced,0.3158880074818929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,64,balanced,0.11009066303571065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,64,balanced,0.11223466197649638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,64,balanced,0.11641066273053487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,64,balanced,0.1332373321056366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,64,balanced,0.3790186643600464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,64,balanced,0.1393226683139801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,64,balanced,0.16156267126401266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,64,balanced,0.18478399515151978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,64,balanced,0.5355146725972494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,64,balanced,0.2545386751492818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,64,balanced,0.2927199999491374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,64,balanced,0.6806346575419108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,64,balanced,0.40404268105824787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,64,balanced,0.5116053422292074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,64,balanced,0.8341120084126791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,64,balanced,0.6231840054194132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,64,balanced,0.029711998999118805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,64,balanced,0.030373332401116688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,64,balanced,0.029029332101345062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,64,balanced,0.03070933371782303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,64,balanced,0.03429333368937174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,64,balanced,0.04266666869322459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,64,balanced,0.06317333380381267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,64,balanced,0.06737066805362701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,64,balanced,0.0718560020128886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,64,balanced,0.07282133400440216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,64,balanced,0.07056533296902974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,64,balanced,0.06443200012048085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,64,balanced,1.2769866784413655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,64,balanced,0.0631520003080368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,64,balanced,0.05271466573079427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,64,balanced,0.08701866865158081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,64,balanced,0.0883840024471283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,64,balanced,0.07205333312352498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,64,balanced,0.0937653382619222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,64,balanced,0.9528373082478842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,64,balanced,0.11338133613268535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,64,balanced,0.15253866712252298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,64,balanced,0.1826080083847046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,64,balanced,0.2630346616109212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,64,balanced,0.33290666341781616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,64,balanced,0.4941759904225667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,64,balanced,0.6572426557540894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,64,balanced,0.3015039960543315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,64,balanced,0.9792799949645996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,64,balanced,0.414522647857666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,64,balanced,2.515157381693522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,64,balanced,1.8570559819539387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,64,balanced,1.2955520153045654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,64,balanced,0.5207573175430298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,64,balanced,1.6071732838948567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,64,balanced,0.6285706758499146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,64,balanced,2.548288027445475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,64,balanced,0.9343893527984619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,64,balanced,5.031776110331218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,64,balanced,1.8215893109639485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,64,power_law_1.01,0.1334720015525818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,64,power_law_1.01,0.12467199563980103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,64,power_law_1.01,0.08451200127601624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,64,power_law_1.01,0.10798720121383668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,64,power_law_1.01,0.0809984028339386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,64,power_law_1.01,0.105132794380188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,64,power_law_1.01,0.0791167974472046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,64,power_law_1.01,0.07305600047111512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,64,power_law_1.01,0.10967680215835571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,64,power_law_1.01,0.08352000117301941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,64,power_law_1.01,0.12165119647979736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,64,power_law_1.01,0.08750720024108886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,64,power_law_1.01,0.1291264057159424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,64,power_law_1.01,0.12749439477920532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,64,power_law_1.01,0.08723840117454529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,64,power_law_1.01,0.09696000218391418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,64,power_law_1.01,0.13216639757156373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,64,power_law_1.01,0.09151999950408936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,64,power_law_1.01,0.13187839984893798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,64,power_law_1.01,0.09715200066566468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,64,power_law_1.01,0.13290879726409913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,64,power_law_1.01,0.1274623990058899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,64,power_law_1.01,0.09587839841842652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,64,power_law_1.01,0.09628159999847412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,64,power_law_1.01,0.12775039672851562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,64,power_law_1.01,0.09763839840888977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,64,power_law_1.01,0.13505280017852783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,64,power_law_1.01,0.1286080002784729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,64,power_law_1.01,0.10135680437088013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,64,power_law_1.01,0.10249600410461426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,64,power_law_1.01,0.13467520475387573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,64,power_law_1.01,0.10417920351028442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,64,power_law_1.01,0.13464959859848022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,64,power_law_1.01,0.10304640531539917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,64,power_law_1.01,0.15667200088500977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,64,power_law_1.01,0.12615679502487182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,64,power_law_1.01,0.18247040510177612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,64,power_law_1.01,0.13975679874420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,64,power_law_1.01,0.18442879915237426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,64,power_law_1.01,0.17134079933166504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,64,power_law_1.01,0.23351678848266602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,64,power_law_1.01,0.20352640151977539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,64,power_law_1.01,0.2805183887481689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,64,power_law_1.01,0.2519295930862427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,64,power_law_1.01,0.3312704086303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,64,power_law_1.01,0.309555196762085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,64,power_law_1.01,0.44864640235900877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,64,power_law_1.01,0.39802238941192625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,64,power_law_1.01,0.5364096164703369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,64,power_law_1.01,0.5268799781799316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,64,power_law_1.01,0.835257625579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,64,power_law_1.01,0.746553611755371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,64,power_law_1.01,0.9849535942077636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,64,power_law_1.01,0.9838912010192871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,64,power_law_1.01,1.1574399948120118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,64,power_law_1.01,1.159500789642334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,64,power_law_1.01,1.9397247314453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,64,power_law_1.01,1.8670528411865235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,64,power_law_1.01,3.8990337371826174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,64,power_law_1.01,3.7998271942138673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,64,power_law_1.01,0.1018623948097229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,64,power_law_1.01,0.0889792025089264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,64,power_law_1.01,0.07774720191955567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,64,power_law_1.01,0.07652480006217957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,64,power_law_1.01,0.08043519854545593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,64,power_law_1.01,0.09196799993515015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,64,power_law_1.01,0.09664639830589294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,64,power_law_1.01,0.09696639776229858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,64,power_law_1.01,0.09238399863243103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,64,power_law_1.01,0.09372159838676453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,64,power_law_1.01,0.09619200229644775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,64,power_law_1.01,0.09360640048980713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,64,power_law_1.01,0.09631999731063842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,64,power_law_1.01,0.09872639775276185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,64,power_law_1.01,0.09681280255317688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,64,power_law_1.01,0.10421760082244873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,64,power_law_1.01,0.02805120050907135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,64,power_law_1.01,0.105075204372406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,64,power_law_1.01,0.028780800104141236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,64,power_law_1.01,0.11205120086669922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,64,power_law_1.01,0.02839680016040802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.01,0.13073279857635497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,64,power_law_1.01,0.03047040104866028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.01,0.14863359928131104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,64,power_law_1.01,0.03410559892654419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.01,0.17422080039978027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,64,power_law_1.01,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.01,0.21210880279541017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,64,power_law_1.01,0.06541439890861511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,64,power_law_1.01,0.06402559876441956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.01,0.26007039546966554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,64,power_law_1.01,0.07722880244255066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.01,0.3339904069900513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,64,power_law_1.01,0.07719680070877075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.01,0.437337589263916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,64,power_law_1.01,0.07410560250282287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.01,0.6118271827697754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,64,power_law_1.01,0.07371519804000855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.01,0.7407807826995849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,64,power_law_1.01,0.07212799787521362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,64,power_law_1.01,0.9889663696289063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,64,power_law_1.01,0.06972159743309021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,64,power_law_1.01,1.5478400230407714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,64,power_law_1.01,0.06774399876594543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,64,power_law_1.01,0.06654719710350036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,64,power_law_1.01,3.0107519149780275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,64,power_law_1.01,0.07546240091323853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,64,power_law_1.01,0.07389439940452576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.01,0.08903679847717286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.01,0.10560640096664428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.01,0.11511039733886719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.01,0.1484544038772583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.01,0.18984960317611693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.01,0.26666879653930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.01,0.35733120441436766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.01,0.5003200054168702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.01,0.6867008209228516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,64,power_law_1.01,0.8851776123046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,64,power_law_1.01,1.3003520011901855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,64,power_law_1.01,2.5979904174804687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,64,power_law_1.2,0.08694400191307068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,64,power_law_1.2,0.12496639490127563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,64,power_law_1.2,0.09008640050888062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,64,power_law_1.2,0.12705279588699342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,64,power_law_1.2,0.07162240147590637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,64,power_law_1.2,0.09438719749450683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,64,power_law_1.2,0.08080000281333924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,64,power_law_1.2,0.10499199628829955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,64,power_law_1.2,0.07975680232048035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,64,power_law_1.2,0.11440000534057618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,64,power_law_1.2,0.08627840280532836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,64,power_law_1.2,0.12566399574279785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,64,power_law_1.2,0.09629439711570739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,64,power_law_1.2,0.12586239576339722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,64,power_law_1.2,0.09414399862289428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,64,power_law_1.2,0.13257600069046022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,64,power_law_1.2,0.09531520009040832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,64,power_law_1.2,0.11969280242919922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,64,power_law_1.2,0.09544960260391236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,64,power_law_1.2,0.13068159818649291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,64,power_law_1.2,0.09562879800796509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,64,power_law_1.2,0.12746880054473878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,64,power_law_1.2,0.09473919868469238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,64,power_law_1.2,0.1309056043624878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,64,power_law_1.2,0.10017280578613282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,64,power_law_1.2,0.1304255962371826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,64,power_law_1.2,0.09633920192718506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,64,power_law_1.2,0.13063679933547973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,64,power_law_1.2,0.10220799446105958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,64,power_law_1.2,0.1313472032546997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,64,power_law_1.2,0.10304000377655029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,64,power_law_1.2,0.13418240547180177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,64,power_law_1.2,0.11304960250854493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,64,power_law_1.2,0.15107840299606323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,64,power_law_1.2,0.134879994392395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,64,power_law_1.2,0.1694208025932312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,64,power_law_1.2,0.15177600383758544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,64,power_law_1.2,0.18255360126495362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,64,power_law_1.2,0.17523839473724365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,64,power_law_1.2,0.20865280628204347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,64,power_law_1.2,0.21984639167785644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,64,power_law_1.2,0.25671679973602296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,64,power_law_1.2,0.27463040351867674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,64,power_law_1.2,0.35808000564575193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,64,power_law_1.2,0.35674879550933836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,64,power_law_1.2,0.4317376136779785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,64,power_law_1.2,0.49514241218566896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,64,power_law_1.2,0.4664576053619385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,64,power_law_1.2,0.7469759941101074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,64,power_law_1.2,0.6147071838378906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,64,power_law_1.2,0.9682687759399414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,64,power_law_1.2,0.890329647064209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,64,power_law_1.2,1.079481601715088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,64,power_law_1.2,1.4842047691345215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,64,power_law_1.2,1.3322175979614257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,64,power_law_1.2,1.9216447830200196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,64,power_law_1.2,2.832691192626953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,64,power_law_1.2,3.269331359863281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,64,power_law_1.2,4.830976104736328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,64,power_law_1.2,5.87146224975586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,64,power_law_1.2,0.10169600248336792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,64,power_law_1.2,0.0884223997592926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,64,power_law_1.2,0.06830080151557923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,64,power_law_1.2,0.08099200129508972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,64,power_law_1.2,0.07967360019683838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,64,power_law_1.2,0.0803391993045807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,64,power_law_1.2,0.09408640265464782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,64,power_law_1.2,0.09050880074501037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,64,power_law_1.2,0.09402239918708802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,64,power_law_1.2,0.09537280201911927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,64,power_law_1.2,0.09207680225372314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,64,power_law_1.2,0.09909120202064514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,64,power_law_1.2,0.027852800488471986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,64,power_law_1.2,0.09632639884948731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,64,power_law_1.2,0.09472640156745911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,64,power_law_1.2,0.029945600032806396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,64,power_law_1.2,0.10165760517120362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,64,power_law_1.2,0.02690559923648834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,64,power_law_1.2,0.10440319776535034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,64,power_law_1.2,0.028255999088287354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,64,power_law_1.2,0.109990394115448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,64,power_law_1.2,0.03544960021972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,64,power_law_1.2,0.13144960403442382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,64,power_law_1.2,0.0453247994184494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.2,0.13685760498046876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,64,power_law_1.2,0.06293119788169861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.2,0.1583359956741333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,64,power_law_1.2,0.061715197563171384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.2,0.18426239490509033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,64,power_law_1.2,0.077811199426651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.2,0.23674240112304687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,64,power_law_1.2,0.07611520290374756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,64,power_law_1.2,0.07460479736328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.2,0.27918078899383547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,64,power_law_1.2,0.0711679995059967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.2,0.4119999885559082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,64,power_law_1.2,0.07236480116844177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.2,0.5089280128479003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,64,power_law_1.2,0.06669440269470214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.2,0.7241343975067138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,64,power_law_1.2,0.0656063973903656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.2,1.0179903984069825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,64,power_law_1.2,0.06929919719696045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,64,power_law_1.2,1.1733887672424317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,64,power_law_1.2,0.07870720028877258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,64,power_law_1.2,1.834886360168457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,64,power_law_1.2,0.07370880246162415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.2,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,64,power_law_1.2,4.43243522644043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.2,0.10302079916000366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.2,0.12579840421676636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.2,0.16629120111465454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.2,0.2101759910583496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.2,0.31466240882873536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.2,0.39130239486694335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.2,0.6204800128936767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.2,0.7839807987213134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,64,power_law_1.2,1.0048640251159668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,32,balanced,0.06634133557478587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,64,power_law_1.2,1.5158464431762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,32,balanced,0.06982933481534322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,32,balanced,0.07036800185839336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,64,power_law_1.2,3.051353645324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,32,balanced,0.08126933375994365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,32,balanced,0.09809600313504536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,32,balanced,0.12156800429026286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,32,balanced,0.12527466813723245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,32,balanced,0.12316266695658366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,32,balanced,0.05482666691144308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,32,balanced,0.1246613363424937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,32,balanced,0.058703998724619545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,32,balanced,0.1267626682917277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,32,balanced,0.056986664732297264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,32,balanced,0.12096533179283142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,32,balanced,0.062277331948280334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,32,balanced,0.0718399981657664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,32,balanced,0.12225066622098286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,32,balanced,0.09145067135492961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,32,balanced,0.12874666849772134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,32,balanced,0.08949866890907288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,32,balanced,0.1256533364454905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,32,balanced,0.09417066971460979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,32,balanced,0.12678933143615723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,32,balanced,0.09230933586756389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,32,balanced,0.09444800019264221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,32,balanced,0.13010133306185404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,32,balanced,0.09558932979901631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,32,balanced,0.12928533554077148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,32,balanced,0.0962666670481364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,32,balanced,0.09634666641553243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,32,balanced,0.1369706690311432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,32,balanced,0.09874666730562846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,32,balanced,0.13590400417645773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,32,balanced,0.10511466860771179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,32,balanced,0.10668800274531047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,32,balanced,0.14941333731015524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,32,balanced,0.10898133118947347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,32,balanced,0.1588479975859324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,32,balanced,0.1167093316713969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,32,balanced,0.1255466639995575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,32,balanced,0.18138132492701212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,32,balanced,0.1455626686414083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,32,balanced,0.16325333714485168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,32,balanced,0.20120533307393393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,32,balanced,0.2007253368695577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,32,balanced,0.2750133275985718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,32,balanced,0.2376586596171061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,32,balanced,0.3088906606038411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,32,balanced,0.3213813304901123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,32,balanced,0.42631999651590985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,32,balanced,0.39027198155721027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,32,balanced,0.5382026831309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,32,balanced,0.5534720023473104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,32,balanced,0.6485013167063395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,32,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,32,balanced,0.05736533304055532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,32,balanced,0.05718400080998739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,32,balanced,0.06366933385531108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,32,balanced,0.08020266890525818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,32,balanced,0.09268800417582194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,32,balanced,0.706607977549235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,32,balanced,0.09809066851933797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,32,balanced,0.09322667121887207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,32,balanced,0.09181333581606548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,32,balanced,0.0942026674747467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,32,balanced,0.09810133775075276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,32,balanced,0.0943999985853831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,32,balanced,0.09386666615804036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,32,balanced,0.10037866234779358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,32,balanced,0.025578667720158894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,32,balanced,0.026965332527955372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,32,balanced,0.10005333026250203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,32,balanced,0.026719999810059864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,32,balanced,0.10626133282979329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,32,balanced,0.026693334182103474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,32,balanced,0.028069332242012024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,32,balanced,0.10545066992441814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,32,balanced,0.03673599908749262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,32,balanced,0.11581333478291829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,32,balanced,0.04483200112978617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,32,balanced,0.9682506720225016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,32,balanced,0.04531733194986979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,32,balanced,0.11529599626859029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,32,balanced,0.06685866912206014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,32,balanced,0.06708799799283345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,32,balanced,0.06784533460934956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,32,balanced,0.1304480036099752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,32,balanced,0.07352533439795177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,32,balanced,0.8669119675954183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,32,balanced,0.07134399811426799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,32,balanced,0.14134400089581808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,32,balanced,0.06515733400980632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,32,balanced,0.09193600217501323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,32,balanced,0.09063466389973958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,32,balanced,0.16705065965652466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,32,balanced,0.08252266546090443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,32,balanced,0.09171733260154724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,32,balanced,0.10797866185506184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,32,balanced,0.19314666589101157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,32,balanced,0.12567999958992004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,32,balanced,0.15500799814860025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,32,balanced,0.2622026602427165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,32,balanced,0.20302400986353555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,32,balanced,0.3046293258666992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,32,balanced,0.2505653301874797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,32,balanced,0.34427201747894287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,32,balanced,1.3305439949035645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,32,balanced,0.42686935265858966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,32,balanced,0.43637335300445557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,32,balanced,1.8954399426778157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,32,balanced,0.5410506725311279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,32,balanced,0.6758453051249186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,32,balanced,0.8780266443888346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,32,balanced,0.6608533461888632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,32,balanced,1.100495974222819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,32,balanced,1.0097119808197021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,32,balanced,2.620698610941569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,32,balanced,1.7201439539591472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,32,balanced,1.9719200134277344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,32,balanced,3.407893180847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,32,power_law_1.01,0.12174719572067261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,32,power_law_1.01,0.11786240339279175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,32,power_law_1.01,0.09376639723777772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,32,power_law_1.01,0.09378560185432434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,32,power_law_1.01,0.10215679407119752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,32,power_law_1.01,0.1072383999824524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,32,power_law_1.01,0.11923199892044067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,32,power_law_1.01,0.11743359565734864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,32,power_law_1.01,0.11904640197753906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,32,power_law_1.01,0.11733119487762451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,32,power_law_1.01,0.1198591947555542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,32,power_law_1.01,0.11936000585556031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,32,power_law_1.01,0.11819519996643066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,32,power_law_1.01,0.12151039838790893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,32,power_law_1.01,0.12064640522003174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,32,power_law_1.01,0.11765120029449463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,32,power_law_1.01,0.1297600030899048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,32,power_law_1.01,0.14052480459213257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,32,power_law_1.01,0.14789119958877564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,32,power_law_1.01,0.18754559755325317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,32,power_law_1.01,0.20170240402221679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,32,power_law_1.01,0.24992640018463136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,32,power_law_1.01,0.2960767984390259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,32,power_law_1.01,0.388319993019104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,32,power_law_1.01,0.48862080574035643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,32,power_law_1.01,0.65381760597229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,32,power_law_1.01,0.8593791961669922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,32,power_law_1.01,1.0530495643615723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,32,power_law_1.01,1.732819175720215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,32,power_law_1.01,3.135257530212402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,32,power_law_1.01,0.09033600091934205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,32,power_law_1.01,0.08772479891777038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,32,power_law_1.01,0.07683839797973632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,32,power_law_1.01,0.06824960112571717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,32,power_law_1.01,0.08048639893531799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,32,power_law_1.01,0.082259202003479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,32,power_law_1.01,0.08938239812850952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,32,power_law_1.01,0.09047679901123047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,32,power_law_1.01,0.08812159895896912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,32,power_law_1.01,0.0895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,32,power_law_1.01,0.09120000004768372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,32,power_law_1.01,0.09104639887809754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,32,power_law_1.01,0.09682559967041016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,32,power_law_1.01,0.09907199740409851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,32,power_law_1.01,0.10209280252456665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,32,power_law_1.01,0.102457594871521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,32,power_law_1.01,0.10901119709014892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,32,power_law_1.01,0.11737600564956666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,32,power_law_1.01,0.1375040054321289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,32,power_law_1.01,0.16391040086746217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,32,power_law_1.01,0.18262399435043336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,32,power_law_1.01,0.23278720378875734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,32,power_law_1.01,0.29456000328063964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,32,power_law_1.01,0.40327038764953616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,32,power_law_1.01,0.10172799825668336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,32,power_law_1.01,0.5025472164154052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,32,power_law_1.01,0.09968640208244324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,32,power_law_1.01,0.7041152000427247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,32,power_law_1.01,0.07844480276107788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,32,power_law_1.01,0.8990079879760742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,32,power_law_1.01,0.06958720088005066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,32,power_law_1.01,1.161030387878418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,32,power_law_1.01,0.08302720189094544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,32,power_law_1.01,1.7155263900756836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,32,power_law_1.01,0.08180480003356934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,32,power_law_1.01,0.09761279821395874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,32,power_law_1.01,3.636857604980469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,32,power_law_1.01,0.08825600147247314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,32,power_law_1.01,0.02744320034980774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,32,power_law_1.01,0.09122560024261475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,32,power_law_1.01,0.027315199375152588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,32,power_law_1.01,0.09731839895248413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,32,power_law_1.01,0.02515200078487396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,32,power_law_1.01,0.0950976014137268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,32,power_law_1.01,0.025561600923538208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,32,power_law_1.01,0.09623680114746094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,32,power_law_1.01,0.027987200021743774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,32,power_law_1.01,0.09742079973220825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,32,power_law_1.01,0.03157120048999786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,32,power_law_1.01,0.09667199850082397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,32,power_law_1.01,0.042905598878860474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,32,power_law_1.01,0.10105600357055664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,32,power_law_1.01,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,32,power_law_1.01,0.10403200387954711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,32,power_law_1.01,0.0603007972240448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,32,power_law_1.01,0.1052672028541565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,32,power_law_1.01,0.06289920210838318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,32,power_law_1.01,0.11938560009002686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,32,power_law_1.01,0.06318079829216003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.01,0.12640639543533325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,32,power_law_1.01,0.07393280267715455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.01,0.14528640508651733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,32,power_law_1.01,0.07612159848213196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.01,0.16506240367889405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,32,power_law_1.01,0.06787840127944947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.01,0.20970239639282226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,32,power_law_1.01,0.07052800059318542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,32,power_law_1.01,0.07054719924926758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.01,0.2510143995285034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,32,power_law_1.01,0.07601280212402343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.01,0.3184959888458252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,32,power_law_1.01,0.07454720139503479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.01,0.4093696117401123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.01,0.08166400194168091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.01,0.584876823425293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.01,0.09112319946289063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.01,0.7488639831542969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.01,0.10189440250396728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,32,power_law_1.01,0.8968319892883301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.01,0.13678079843521118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,32,power_law_1.01,1.363225555419922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.01,0.15875200033187867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,32,power_law_1.01,2.8569536209106445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.01,0.2164544105529785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.01,0.26617600917816164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.01,0.3919167995452881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.01,0.5350272178649902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,32,power_law_1.01,0.651039981842041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,32,power_law_1.01,0.9824511528015136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,32,power_law_1.01,2.0077888488769533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,32,power_law_1.2,0.12671999931335448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,32,power_law_1.2,0.12319359779357911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,32,power_law_1.2,0.07920640110969543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,32,power_law_1.2,0.09121919870376587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,32,power_law_1.2,0.09604480266571044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,32,power_law_1.2,0.10488959550857543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,32,power_law_1.2,0.11134079694747925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,32,power_law_1.2,0.11683199405670167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,32,power_law_1.2,0.1218176007270813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,32,power_law_1.2,0.11784960031509399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,32,power_law_1.2,0.11816960573196411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,32,power_law_1.2,0.11772799491882324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,32,power_law_1.2,0.11681280136108399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,32,power_law_1.2,0.11544320583343506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,32,power_law_1.2,0.12677760124206544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,32,power_law_1.2,0.12311040163040161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,32,power_law_1.2,0.13649280071258546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,32,power_law_1.2,0.15168639421463012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,32,power_law_1.2,0.1668352007865906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,32,power_law_1.2,0.1926144003868103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,32,power_law_1.2,0.21250560283660888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,32,power_law_1.2,0.2742144107818604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,32,power_law_1.2,0.3524735927581787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,32,power_law_1.2,0.4393599987030029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,32,power_law_1.2,0.5629248142242431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,32,power_law_1.2,0.8171839714050293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,32,power_law_1.2,1.0536767959594726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,32,power_law_1.2,1.3209088325500489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,32,power_law_1.2,2.0631359100341795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,32,power_law_1.2,4.408524703979492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,32,power_law_1.2,0.08966400027275086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,32,power_law_1.2,0.08576639890670776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,32,power_law_1.2,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,32,power_law_1.2,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,32,power_law_1.2,0.07549440264701843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,32,power_law_1.2,0.07872639894485474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,32,power_law_1.2,0.08650239706039428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,32,power_law_1.2,0.09175040125846863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,32,power_law_1.2,0.09260799884796142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,32,power_law_1.2,0.08837760090827942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,32,power_law_1.2,0.08991360068321227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,32,power_law_1.2,0.09326720237731934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,32,power_law_1.2,0.09422720074653626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,32,power_law_1.2,0.09776639938354492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,32,power_law_1.2,0.10202879905700683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,32,power_law_1.2,0.10629760026931763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,32,power_law_1.2,0.1109887957572937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,32,power_law_1.2,0.10069760084152221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,32,power_law_1.2,0.12887680530548096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,32,power_law_1.2,0.0980351984500885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,32,power_law_1.2,0.14422399997711183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,32,power_law_1.2,0.06307200193405152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,32,power_law_1.2,0.16660480499267577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,32,power_law_1.2,0.0722495973110199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,32,power_law_1.2,0.20397439002990722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,32,power_law_1.2,0.07215359807014465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,32,power_law_1.2,0.25626239776611326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,32,power_law_1.2,0.07858560085296631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,32,power_law_1.2,0.3107264041900635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,32,power_law_1.2,0.08860160112380981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,32,power_law_1.2,0.4332672119140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,32,power_law_1.2,0.08648319840431214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,32,power_law_1.2,0.6158143997192382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,32,power_law_1.2,0.09377279877662659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,32,power_law_1.2,0.8085696220397949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,32,power_law_1.2,0.027673599123954774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,32,power_law_1.2,0.09360640048980713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,32,power_law_1.2,1.0394111633300782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,32,power_law_1.2,0.02768000066280365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,32,power_law_1.2,0.09337599873542786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,32,power_law_1.2,1.1950271606445313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,32,power_law_1.2,0.024249599874019624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,32,power_law_1.2,0.09196799993515015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,32,power_law_1.2,1.8909952163696289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,32,power_law_1.2,0.024447999894618988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,32,power_law_1.2,0.09381759762763978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,32,power_law_1.2,0.09499520063400269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,32,power_law_1.2,0.027347201108932497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,32,power_law_1.2,4.494137573242187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,32,power_law_1.2,0.10160000324249267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,32,power_law_1.2,0.030553600192070006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,32,power_law_1.2,0.10400639772415161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,32,power_law_1.2,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,32,power_law_1.2,0.113811194896698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,32,power_law_1.2,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,32,power_law_1.2,0.12168320417404174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,32,power_law_1.2,0.061817598342895505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.2,0.13583999872207642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,32,power_law_1.2,0.06253439784049988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.2,0.15427199602127076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,32,power_law_1.2,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.2,0.18161280155181886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,32,power_law_1.2,0.07201279997825623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,32,power_law_1.2,0.0726144015789032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.2,0.22256639003753662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,32,power_law_1.2,0.06935679912567139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.2,0.2832832098007202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,32,power_law_1.2,0.07195519804954528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,32,power_law_1.2,0.06704000234603882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,32,power_law_1.2,0.07495679855346679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,32,power_law_1.2,0.07648640275001525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.2,0.08255360126495362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.2,0.08916479945182801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.2,0.09815040230751038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.2,0.1424191951751709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.2,0.16191359758377075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.2,0.2350208044052124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.2,0.29062399864196775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.2,0.4443391799926758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.2,0.5385151863098144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,32,power_law_1.2,0.7299647808074952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,32,power_law_1.2,1.1493632316589355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,32,power_law_1.2,2.5167999267578125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,16,balanced,0.058517331878344216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,16,balanced,0.06100266675154368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,16,balanced,0.06354133288065593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,16,balanced,0.0791733314593633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,16,balanced,0.0926080048084259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,16,balanced,0.11866133411725362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,16,balanced,0.1249066690603892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,16,balanced,0.12027733524640401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,16,balanced,0.12692266702651978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,16,balanced,0.12369599938392639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,16,balanced,0.11705600221951802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,16,balanced,0.11723732948303223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,16,balanced,0.12483200430870056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,16,balanced,0.11958400408426921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,16,balanced,0.12329066793123881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,16,balanced,0.1274666686852773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,16,balanced,0.12809066971143088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,16,balanced,0.1365120013554891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,16,balanced,0.14090133706728616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,16,balanced,0.15709333618481955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,16,balanced,0.16658133268356323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,16,balanced,0.19124799966812134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,16,balanced,0.21249600251515707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,16,balanced,0.28966933488845825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,16,balanced,0.33284799257914227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,16,balanced,0.46374932924906415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,16,balanced,0.5873920122782389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,16,balanced,0.05269866685072581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,16,balanced,0.0531626691420873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,16,balanced,0.05572799841562907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,16,balanced,0.06182933350404104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,16,balanced,0.07292800148328145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,16,balanced,0.09327466289202373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,16,balanced,0.09460799892743428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,16,balanced,0.09217600027720134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,16,balanced,0.09338133533795674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,16,balanced,0.7068586349487305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,16,balanced,0.09570133686065674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,16,balanced,0.09542399644851685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,16,balanced,0.09612799684206645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,16,balanced,0.09727999567985535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,16,balanced,0.09808533390363057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,16,balanced,0.10408533612887065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,16,balanced,0.10596799850463867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,16,balanced,0.11065600315729777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,16,balanced,0.12343999743461609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,16,balanced,0.13026133179664612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,16,balanced,0.1523306667804718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,16,balanced,0.169429341952006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.2,0.39456000328063967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.2,0.4693439960479736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,16,balanced,0.20974934101104736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,16,balanced,1.0685386657714844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.2,0.6920383930206299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.2,0.8192576408386231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,16,balanced,0.0517546683549881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,16,balanced,0.053210665782292686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,32,power_law_1.2,1.1859904289245606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,16,balanced,0.24771199623743692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,16,balanced,0.05745600163936615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,16,balanced,0.06344000001748402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,32,power_law_1.2,1.7853631973266602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,16,balanced,0.07121600210666656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,16,balanced,0.0983679989973704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,32,power_law_1.2,3.6126529693603517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,16,balanced,0.09828799962997437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,16,balanced,0.10179199775060017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,16,balanced,0.3438719908396403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,16,balanced,0.10264000296592712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,16,balanced,0.10013332962989807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,16,balanced,0.09619200229644775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,16,balanced,0.09562666217486064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,16,balanced,0.10195733110109965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,16,balanced,0.4204373359680176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,16,balanced,0.10085333387056987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,16,balanced,0.10451733072598775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,16,balanced,0.11256532867749532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,16,balanced,0.10995733737945557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,16,balanced,0.1165066659450531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,16,balanced,0.1264906624952952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,16,balanced,0.5964106718699137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,16,balanced,0.13988799850145975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,16,balanced,0.15122133493423462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,16,balanced,0.1816800038019816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,16,balanced,0.20673600832621256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,16,balanced,2.088688055674235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,16,balanced,0.7710293134053549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,16,balanced,0.28573334217071533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,16,balanced,0.3410026629765828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,16,balanced,0.9416906833648682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,16,balanced,0.47300267219543457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,16,balanced,0.6052159865697225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,16,balanced,1.4443359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,16,balanced,0.7414399782816569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,16,balanced,1.1417120297749836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,16,balanced,0.02382933348417282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,16,balanced,0.025066666305065155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,16,balanced,0.026613332331180573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,16,balanced,0.026378666361172993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,16,balanced,0.028368001182874043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,16,balanced,0.036389333506425224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,16,balanced,0.04260266820589701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,16,balanced,0.04206933577855428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,16,balanced,0.05491200089454651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,16,balanced,0.055455997586250305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,16,balanced,0.05497066676616669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,16,balanced,2.8409598668416343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,16,balanced,0.09261332949002583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,16,balanced,0.09198932846387227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,16,balanced,0.09153067072232564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,16,balanced,0.09040533502896626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,16,balanced,0.08970133463541667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,16,power_law_1.01,0.09737600088119507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,16,balanced,0.08734933535257976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,16,power_law_1.01,0.12378239631652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,16,power_law_1.01,0.08460800051689148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,16,balanced,0.0939520001411438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,16,power_law_1.01,0.0818560004234314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,16,balanced,0.10221866766611735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,16,power_law_1.01,0.08983039855957031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,16,power_law_1.01,0.09854080080986023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,16,balanced,0.12180800239245097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,16,power_law_1.01,0.11401599645614624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,16,power_law_1.01,0.10941439867019653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,16,power_law_1.01,0.10813440084457397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,16,balanced,0.14486400286356607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,16,power_law_1.01,0.10937600135803223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,16,power_law_1.01,0.11194239854812622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,16,balanced,0.17671465873718262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,16,power_law_1.01,0.11325440406799317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,16,power_law_1.01,0.11552000045776367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,16,power_law_1.01,0.12112640142440796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,16,power_law_1.01,0.12412159442901612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,16,balanced,0.21434134244918823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,16,power_law_1.01,0.12491519451141357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,16,power_law_1.01,0.12732160091400146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,16,balanced,2.2295467058817544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,16,power_law_1.01,0.13378560543060303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,16,power_law_1.01,0.15121920108795167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,16,power_law_1.01,0.18591359853744507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,16,balanced,0.2759679953257243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,16,power_law_1.01,0.19720959663391113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,16,power_law_1.01,0.2395263910293579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,16,power_law_1.01,0.2926271915435791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,16,power_law_1.01,0.3496959924697876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,16,power_law_1.01,0.47222399711608887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,16,power_law_1.01,0.644371223449707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,16,balanced,0.34110931555430096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,16,power_law_1.01,0.7993216037750244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,16,power_law_1.01,1.0426688194274902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,16,power_law_1.01,1.5401535987854005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,16,power_law_1.01,3.3723072052001952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,16,balanced,0.5162453254063925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,16,balanced,0.671674648920695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,16,balanced,0.8545973300933838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,16,power_law_1.01,0.07359359860420227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,16,power_law_1.01,0.09548159837722778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,16,power_law_1.01,0.06863999962806702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,16,power_law_1.01,0.06458879709243774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,16,power_law_1.01,0.07539839744567871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,16,power_law_1.01,0.07964159846305847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,16,power_law_1.01,0.09044479727745056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,16,power_law_1.01,0.08766720294952393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,16,power_law_1.01,0.09022079706192017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,16,power_law_1.01,0.08886399865150452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,16,power_law_1.01,0.0897599995136261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,16,balanced,1.3135146299997966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,16,power_law_1.01,0.09266560077667237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,16,power_law_1.01,0.09515519738197327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,16,power_law_1.01,0.09747200012207032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,16,power_law_1.01,0.1024448037147522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,16,power_law_1.01,0.10540159940719604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,16,power_law_1.01,0.11066240072250366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,16,power_law_1.01,0.1240447998046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,16,power_law_1.01,0.1396607995033264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,16,power_law_1.01,0.17288960218429567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,16,power_law_1.01,0.19345279932022094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,16,power_law_1.01,0.24537599086761475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,16,power_law_1.01,0.07984640002250672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,16,power_law_1.01,0.3041663885116577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,16,power_law_1.01,0.10266879796981812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,16,power_law_1.01,0.4065663814544678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,16,power_law_1.01,0.07498239874839782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,16,power_law_1.01,0.510591983795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,16,power_law_1.01,0.06897280216217042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,16,power_law_1.01,0.7805183887481689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,16,power_law_1.01,0.07868800163269044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,16,power_law_1.01,0.9175168037414551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,16,power_law_1.01,0.08220800161361694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,16,power_law_1.01,0.09516159892082214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,16,power_law_1.01,1.183788776397705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,16,power_law_1.01,0.09591680169105529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,16,power_law_1.01,1.7748735427856446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,16,power_law_1.01,0.09768959879875183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,16,power_law_1.01,0.09671040177345276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,16,power_law_1.01,3.863507080078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,16,power_law_1.01,0.09514880180358887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,16,power_law_1.01,0.09994239807128906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,16,power_law_1.01,0.02285439968109131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,16,power_law_1.01,0.10088959932327271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,16,power_law_1.01,0.028569599986076354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,16,power_law_1.01,0.10154240131378174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,16,power_law_1.01,0.023289600014686586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,16,power_law_1.01,0.10566400289535523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,16,power_law_1.01,0.024748800694942473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,16,power_law_1.01,0.10980479717254639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,16,power_law_1.01,0.026956799626350402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,16,power_law_1.01,0.11325440406799317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,16,power_law_1.01,0.030393600463867188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,16,power_law_1.01,0.1229375958442688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,16,power_law_1.01,0.03386240005493164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.01,0.130239999294281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,16,power_law_1.01,0.035667198896408084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.01,0.15182080268859863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,16,power_law_1.01,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.01,0.1735103964805603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,16,power_law_1.01,0.04657280147075653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.01,0.21985280513763428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,16,power_law_1.01,0.04602240025997162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,16,power_law_1.01,0.066457599401474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,16,power_law_1.01,0.0648576021194458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,16,balanced,2.5796213150024414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,16,power_law_1.01,0.06768640279769897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,16,power_law_1.01,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,16,power_law_1.01,0.07904639840126038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,16,power_law_1.01,0.07963520288467407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,16,power_law_1.01,0.07881600260734559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.01,0.08110079765319825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.01,0.0949184000492096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.01,0.10760960578918458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.01,0.12257920503616333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.01,0.14341119527816773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.01,0.2004159927368164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.01,0.2499392032623291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.01,0.34963839054107665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.01,0.4424448013305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,16,power_law_1.01,0.5762688159942627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,16,power_law_1.01,0.846668815612793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,16,power_law_1.01,1.766636848449707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,16,power_law_1.2,0.09440640211105347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,16,power_law_1.2,0.12355200052261353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,16,power_law_1.2,0.07468159794807434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,16,power_law_1.2,0.08041599988937378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,16,power_law_1.2,0.08848639726638793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,16,power_law_1.2,0.0961471974849701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,16,power_law_1.2,0.11212799549102784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,16,power_law_1.2,0.10924160480499268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,16,power_law_1.2,0.11323519945144653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,16,power_law_1.2,0.11130880117416382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,16,power_law_1.2,0.11374080181121826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,16,power_law_1.2,0.11762559413909912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,16,power_law_1.2,0.11558400392532349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,16,power_law_1.2,0.11920640468597413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,16,power_law_1.2,0.12127360105514526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,16,power_law_1.2,0.07285760045051574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,16,power_law_1.2,0.12026879787445069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,16,power_law_1.2,0.08978559970855712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,16,power_law_1.2,0.12963839769363403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,16,power_law_1.2,0.06061440110206604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,16,power_law_1.2,0.1434623956680298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,16,power_law_1.2,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,16,power_law_1.2,0.15896960496902465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,16,power_law_1.2,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,16,power_law_1.2,0.19633920192718507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,16,power_law_1.2,0.07691519856452941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,16,power_law_1.2,0.21338880062103271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,16,power_law_1.2,0.08212479948997498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,16,power_law_1.2,0.2748608112335205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,16,power_law_1.2,0.08770560026168824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,16,power_law_1.2,0.3141439914703369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,16,power_law_1.2,0.08727679848670959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,16,power_law_1.2,0.4116991996765137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,16,power_law_1.2,0.08858240246772767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,16,power_law_1.2,0.5140607833862305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,16,power_law_1.2,0.09034240245819092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,16,power_law_1.2,0.7832255840301514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,16,power_law_1.2,0.0915008008480072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.01,0.2654592037200928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,16,power_law_1.2,1.0766655921936035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,16,power_law_1.2,0.09307519793510437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.01,0.35382399559020994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,16,power_law_1.2,1.142745590209961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,16,power_law_1.2,0.09625599980354309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.01,0.4421504020690918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,16,power_law_1.2,0.10316799879074097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,16,power_law_1.2,1.7281984329223632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.01,0.6151103973388672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,16,power_law_1.2,0.10572799444198608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,16,power_law_1.2,0.07890560030937195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.01,0.7851071834564209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,16,power_law_1.2,3.774790573120117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,16,power_law_1.2,0.11398400068283081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,16,power_law_1.2,0.10490239858627319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,16,power_law_1.01,1.0033727645874024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,16,power_law_1.2,0.1312064051628113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,16,power_law_1.2,0.06775040030479432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,16,power_law_1.01,1.456934356689453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,16,power_law_1.2,0.1412927985191345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,16,power_law_1.2,0.06796159744262695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,16,power_law_1.2,0.1779711961746216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,16,power_law_1.01,2.860723114013672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,16,power_law_1.2,0.07342720031738281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,16,power_law_1.2,0.1990272045135498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,16,power_law_1.2,0.08053759932518005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,16,power_law_1.2,0.2531327962875366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,16,power_law_1.2,0.09442560076713562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,16,power_law_1.2,0.3138495922088623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,16,power_law_1.2,0.09127680063247681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,16,power_law_1.2,0.4472511768341064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,16,power_law_1.2,0.09287679791450501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,16,power_law_1.2,0.5558464050292968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,16,power_law_1.2,0.1018623948097229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,16,power_law_1.2,0.0942463994026184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,16,power_law_1.2,0.757203197479248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,16,power_law_1.2,0.09632639884948731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,16,power_law_1.2,0.9768896102905273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,16,power_law_1.2,0.10026240348815918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,16,power_law_1.2,1.230527973175049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,16,power_law_1.2,0.10252799987792968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,16,power_law_1.2,2.009811210632324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,16,power_law_1.2,0.10842239856719971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,16,power_law_1.2,0.10578559637069702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,16,power_law_1.2,3.9648704528808594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,16,power_law_1.2,0.11546239852905274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,16,power_law_1.2,0.12725759744644166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.2,0.13841919898986815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.2,0.15735039710998536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.2,0.1927616000175476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.2,0.23055999279022216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.2,0.29241600036621096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.2,0.38401920795440675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.2,0.4869120121002197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.2,0.7016704082489014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.2,0.9387711524963379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,16,power_law_1.2,1.130617618560791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,16,power_law_1.2,1.6847808837890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,16,power_law_1.2,3.629715347290039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,16,power_law_1.2,0.023020799458026885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,8,balanced,0.05646933118502299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,16,power_law_1.2,0.02784000039100647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,8,balanced,0.059445331494013466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,8,balanced,0.06339199841022491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,16,power_law_1.2,0.023263999819755556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,8,balanced,0.07805333534876506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,16,power_law_1.2,0.02399359941482544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,8,balanced,0.09617599844932556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,8,balanced,0.12755200266838074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,16,power_law_1.2,0.02710399925708771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,8,balanced,0.12403200070063274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,16,power_law_1.2,0.02961280047893524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,8,balanced,0.1251200040181478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,8,balanced,0.12500799695650736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,16,power_law_1.2,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,8,balanced,0.12553067008654276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,16,power_law_1.2,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,8,balanced,0.12078932921091716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,16,power_law_1.2,0.0457152009010315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,8,balanced,0.1285653313000997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,8,balanced,0.05022933085759481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,16,power_law_1.2,0.04416640102863312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,8,balanced,0.1253546675046285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,8,balanced,0.054101333022117615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,8,balanced,0.12128000458081563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,16,power_law_1.2,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,8,balanced,0.05675200124581655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,8,balanced,0.05942399799823761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,8,balanced,0.12882133324941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,16,power_law_1.2,0.06474879980087281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,8,balanced,0.07715733349323273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,8,balanced,0.128330667813619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,16,power_law_1.2,0.06440960168838501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,8,balanced,0.10282133022944133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,8,balanced,0.10472533106803894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,8,balanced,0.13446933031082153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,16,power_law_1.2,0.06781439781188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,8,balanced,0.10222400228182475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,8,balanced,0.14274666706720987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,16,power_law_1.2,0.0800320029258728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,8,balanced,0.1034346620241801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,8,balanced,0.10311466455459595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,16,power_law_1.2,0.07888000011444092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,8,balanced,0.15009599924087524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,8,balanced,0.1037546694278717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,16,power_law_1.2,0.07833600044250488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,8,balanced,0.10327466328938802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,8,balanced,0.16662933429082236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,16,power_law_1.2,0.07879040241241456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,8,balanced,0.10521066188812256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.2,0.08366720080375671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,8,balanced,0.10802132884661357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,8,balanced,0.18156800667444864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,8,balanced,0.1144533356030782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.2,0.09698560237884521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,8,balanced,0.1148373285929362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.2,0.10597120523452759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,8,balanced,0.12247999509175618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.2,0.13441280126571656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,8,balanced,0.21251734097798666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,8,balanced,0.13411200046539307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.2,0.15708800554275512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.2,0.20407040119171144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,8,balanced,0.1434986690680186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.2,0.27628800868988035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,8,balanced,0.24308266242345175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,8,balanced,0.16845866044362387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.2,0.3864255905151367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.2,0.48341121673583987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,8,balanced,0.18940265973409018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,8,balanced,0.05273066461086273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,16,power_law_1.2,0.5985663890838623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,8,balanced,0.05906666815280914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,8,balanced,0.33905065059661865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,8,balanced,0.05807999769846598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,16,power_law_1.2,0.930726432800293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,8,balanced,0.06903466582298279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,8,balanced,0.2355253299077352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,8,balanced,0.0823466678460439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,16,power_law_1.2,1.916703987121582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,8,balanced,0.11648533741633098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,8,balanced,0.11086400349934895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,8,balanced,0.11768000324567159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,8,balanced,0.11540266871452332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,8,balanced,0.27990933259328205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,8,balanced,0.3952853282292684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,8,balanced,0.11502933502197266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,8,balanced,0.11744532982508342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,8,balanced,0.11345600088437398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,8,balanced,0.11342400312423706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,8,balanced,0.3959093491236369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,8,balanced,0.11899733543395996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,8,balanced,0.12029332915941875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,8,balanced,0.1243893305460612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,8,balanced,0.555893341700236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,8,balanced,0.12820800145467123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,8,balanced,0.13707199692726135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,8,balanced,0.48526934782663983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,8,balanced,0.14429866274197897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,8,balanced,0.16270400087038675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,8,balanced,0.17761067549387613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,8,balanced,0.697322686513265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,8,balanced,0.702560027440389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,8,balanced,0.2129653294881185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,8,balanced,0.24591465791066489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,8,balanced,0.35028799374898273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,8,balanced,0.8957866827646891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,8,balanced,0.8574613730112711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,8,balanced,0.41309332847595215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,8,balanced,1.1008373101552327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,8,balanced,0.5899360179901123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,8,balanced,1.294378678003947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,8,balanced,0.75436798731486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,8,balanced,1.692522684733073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,8,balanced,0.9211359818776449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,8,balanced,1.4306453069051106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,8,balanced,2.529978593190511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,8,balanced,0.02351466566324234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,8,balanced,0.02475200096766154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,8,balanced,0.026874666412671406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,8,balanced,3.3338934580485025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,8,balanced,0.028223998844623566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,8,balanced,0.030645333230495453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,8,balanced,0.040474665661652885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,8,balanced,0.0405173326532046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,8,balanced,0.04274666806062063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,8,balanced,0.046442667643229164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,8,balanced,0.04706133405367533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,8,balanced,0.04849599798520406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,8,balanced,0.06657599906126659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,8,balanced,0.06087466577688853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,8,balanced,0.06323199967543285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,8,balanced,0.09640533725420634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,8,balanced,0.09704533219337463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,8,balanced,0.10000532865524292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,8,balanced,0.10651733477910359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,8,balanced,0.11348266402880351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,8,balanced,0.13728533188501993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,8,balanced,0.14537066221237183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,8,balanced,0.18150933583577475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,8,balanced,2.7957067489624023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,8,balanced,0.20206934213638306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,8,balanced,0.25280000766118366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,8,balanced,0.30714666843414307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,8,balanced,0.4604533513387044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,8,balanced,0.5762186845143636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,8,balanced,0.7376746336619059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,8,power_law_1.01,0.0819263994693756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,8,power_law_1.01,0.0673088014125824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,8,power_law_1.01,0.09827200174331666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,8,power_law_1.01,0.07953919768333435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,8,power_law_1.01,0.07511039972305297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,8,power_law_1.01,0.06808320283889771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,8,power_law_1.01,0.07952640056610108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,8,power_law_1.01,0.06921600103378296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,8,power_law_1.01,0.09200639724731445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,8,power_law_1.01,0.07653120160102844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,8,power_law_1.01,0.09996799826622009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,8,power_law_1.01,0.0818880021572113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,8,power_law_1.01,0.10841599702835084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,8,power_law_1.01,0.09020159840583801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,8,power_law_1.01,0.11534719467163086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,8,power_law_1.01,0.09574400186538697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,8,power_law_1.01,0.11424640417099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,8,power_law_1.01,0.09445760250091553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,8,power_law_1.01,0.1144320011138916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,8,power_law_1.01,0.09596800208091735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,8,power_law_1.01,0.11716480255126953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,8,power_law_1.01,0.09713919758796692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,8,power_law_1.01,0.1162559986114502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,8,power_law_1.01,0.09967359900474548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,8,power_law_1.01,0.11605119705200195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,8,power_law_1.01,0.10216959714889526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,8,power_law_1.01,0.11827839612960815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,8,power_law_1.01,0.10724480152130127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,8,power_law_1.01,0.12636159658432006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,8,power_law_1.01,0.11329920291900634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,8,power_law_1.01,0.12845439910888673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,8,power_law_1.01,0.11424640417099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,8,power_law_1.01,0.13464959859848022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,8,power_law_1.01,0.12153600454330445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,8,power_law_1.01,0.1439296007156372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,8,power_law_1.01,0.13672959804534912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,8,power_law_1.01,0.1606719970703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,8,power_law_1.01,0.15187840461730956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,8,power_law_1.01,0.19156479835510254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,8,balanced,1.1250240008036296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,8,power_law_1.01,0.07176960110664368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,8,power_law_1.01,0.17987200021743774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,8,power_law_1.01,0.21687040328979493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,8,power_law_1.01,0.08824319839477539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,8,power_law_1.01,0.2206399917602539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,8,power_law_1.01,0.2663680076599121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,8,power_law_1.01,0.06250879764556885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,8,power_law_1.01,0.272160005569458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,8,power_law_1.01,0.31677439212799074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,8,power_law_1.01,0.06948480010032654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,8,power_law_1.01,0.33635199069976807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,8,power_law_1.01,0.42220158576965333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,8,power_law_1.01,0.07725440263748169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,8,power_law_1.01,0.4433728218078613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,8,power_law_1.01,0.516864013671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,8,power_law_1.01,0.08855040073394775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,8,power_law_1.01,0.5768832206726074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,8,power_law_1.01,0.7492671966552734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,8,power_law_1.01,0.10180480480194092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,8,power_law_1.01,0.7978816032409668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,8,power_law_1.01,0.8689087867736817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,8,power_law_1.01,0.10394879579544067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,8,power_law_1.01,1.0506752014160157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,8,power_law_1.01,1.0635647773742676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,8,power_law_1.01,0.10473599433898925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,8,power_law_1.01,1.3259648323059081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,8,power_law_1.01,1.7455104827880858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,8,power_law_1.01,0.10744320154190064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,8,power_law_1.01,2.007699203491211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,8,power_law_1.01,0.11016319990158081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,8,power_law_1.01,3.533599853515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,8,power_law_1.01,0.1087231993675232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,8,power_law_1.01,4.114636611938477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,8,power_law_1.01,0.11119359731674194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,8,power_law_1.01,0.11463040113449097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,8,power_law_1.01,0.12360320091247559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,8,power_law_1.01,0.12035839557647705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,8,power_law_1.01,0.1275904059410095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,8,power_law_1.01,0.13968640565872192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.01,0.15021439790725707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.01,0.1810495972633362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.01,0.2040895938873291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.01,0.24316799640655518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.01,0.3096640110015869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.01,0.3965951919555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.01,0.5265984058380127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.01,0.773305606842041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.01,0.9153471946716308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,8,power_law_1.01,1.0949312210083009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,8,power_law_1.01,1.7538175582885742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,8,power_law_1.01,3.5196414947509767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,8,balanced,2.2095999717712402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,8,power_law_1.2,0.07926399707794189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,8,power_law_1.2,0.0942911982536316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,8,power_law_1.2,0.0707264006137848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,8,power_law_1.2,0.08049920201301575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,8,power_law_1.01,0.02176000028848648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,8,power_law_1.2,0.08960639834403991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,8,power_law_1.01,0.02409600019454956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,8,power_law_1.2,0.09411839842796325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,8,power_law_1.01,0.02423039972782135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,8,power_law_1.2,0.10652799606323242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,8,power_law_1.01,0.02584959864616394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,8,power_law_1.2,0.11354240179061889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,8,power_law_1.01,0.02824319899082184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,8,power_law_1.2,0.11103999614715576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,8,power_law_1.01,0.030694401264190672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,8,power_law_1.2,0.11325440406799317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,8,power_law_1.2,0.11101440191268921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,8,power_law_1.01,0.03374080061912536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,8,power_law_1.2,0.11585279703140258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,8,power_law_1.01,0.033907198905944826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,8,power_law_1.2,0.11724799871444702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,8,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,8,power_law_1.2,0.11848959922790528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,8,power_law_1.01,0.03782399892807007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,8,power_law_1.2,0.13100160360336305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,8,power_law_1.01,0.03918719887733459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,8,power_law_1.2,0.13089280128479003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,8,power_law_1.01,0.04938240051269531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,8,power_law_1.2,0.13779200315475465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,8,power_law_1.01,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,8,power_law_1.2,0.15050239562988282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,8,power_law_1.01,0.05051519870758057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,8,power_law_1.2,0.16658560037612916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,8,power_law_1.01,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,8,power_law_1.2,0.2060863971710205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,8,power_law_1.01,0.07314559817314148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,8,power_law_1.2,0.22190721035003663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,8,power_law_1.01,0.07530879974365234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,8,power_law_1.2,0.27023360729217527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,8,power_law_1.01,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,8,power_law_1.2,0.06626560091972351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,8,power_law_1.2,0.3322688102722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,8,power_law_1.2,0.07831040024757385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.01,0.0900223970413208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,8,power_law_1.2,0.4246272087097168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,8,power_law_1.2,0.061427199840545656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.01,0.10104960203170776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,8,power_law_1.2,0.06502400040626526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,8,power_law_1.2,0.5397439956665039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.01,0.1120576024055481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,8,power_law_1.2,0.0732479989528656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,8,power_law_1.2,0.7452608108520508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.01,0.1323583960533142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,8,power_law_1.2,0.08271999955177307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.01,0.14546560049057006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,8,power_law_1.2,0.9618816375732422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,8,power_law_1.2,0.09119359850883484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.01,0.2036288022994995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,8,power_law_1.2,1.1092351913452148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,8,power_law_1.2,0.09470080137252808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.01,0.24247679710388184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,8,power_law_1.2,1.9754816055297852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,8,power_law_1.2,0.09750400185585022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.01,0.3415679931640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,8,power_law_1.2,0.07152640223503112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,8,power_law_1.2,0.09783679842948914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,8,power_law_1.2,4.076544189453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.01,0.43753600120544434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,8,power_law_1.2,0.08456959724426269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,8,power_law_1.2,0.10193920135498047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,8,power_law_1.2,0.060838401317596436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,8,power_law_1.01,0.5264256000518799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,8,power_law_1.2,0.10202239751815796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,8,power_law_1.2,0.06943359971046448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,8,power_law_1.2,0.10028799772262573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,8,power_law_1.01,0.7921855926513672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,8,power_law_1.2,0.07642239928245545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,8,power_law_1.2,0.10540159940719604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,8,power_law_1.2,0.08305280208587647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,8,power_law_1.01,1.6505407333374023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,8,power_law_1.2,0.11396479606628418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,8,power_law_1.2,0.10105600357055664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,8,power_law_1.2,0.11600639820098876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,8,power_law_1.2,0.10410239696502685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,8,power_law_1.2,0.12353919744491577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,8,power_law_1.2,0.10191359519958496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,8,power_law_1.2,0.14353920221328736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,8,power_law_1.2,0.10516480207443238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,8,power_law_1.2,0.15418879985809325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,8,power_law_1.2,0.1133504033088684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,8,power_law_1.2,0.19301120042800904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,8,power_law_1.2,0.11112320423126221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,8,power_law_1.2,0.2179840087890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,8,power_law_1.2,0.11384320259094238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,8,power_law_1.2,0.2850624084472656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,8,power_law_1.2,0.11596159934997559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,8,power_law_1.2,0.3468287944793701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,8,power_law_1.2,0.12224639654159546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,8,power_law_1.2,0.5109055995941162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,8,power_law_1.2,0.12461440563201905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,8,power_law_1.2,0.6005248069763184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,8,power_law_1.2,0.12995840311050416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,8,power_law_1.2,0.8723199844360352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,8,power_law_1.2,0.14566400051116943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,8,power_law_1.2,1.2126336097717285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.2,0.15134079456329347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,8,power_law_1.2,1.5356032371520996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.2,0.1874943971633911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,8,power_law_1.2,2.120070457458496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.2,0.2125823974609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.2,0.27319679260253904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,8,power_law_1.2,4.552934265136718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.2,0.32325758934021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.2,0.42522878646850587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.2,0.5683712005615235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.2,0.7582655906677246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.2,0.9827263832092286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,8,power_law_1.2,1.2844223976135254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,8,power_law_1.2,1.9206592559814453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,8,power_law_1.2,3.949817657470703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,4,balanced,0.05995733539263407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,4,balanced,0.06353066861629486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,4,balanced,0.06516799827416737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,4,balanced,0.0849120020866394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,4,balanced,0.10947733124097188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,8,power_law_1.2,0.02157440036535263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,4,balanced,0.1415893336137136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,8,power_law_1.2,0.0236735999584198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,4,balanced,0.14181333780288696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,8,power_law_1.2,0.023820799589157105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,4,balanced,0.14145066340764365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,4,balanced,0.14411733547846475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,8,power_law_1.2,0.02479359954595566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,4,balanced,0.1409280002117157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,8,power_law_1.2,0.028377598524093627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,4,balanced,0.1383946637312571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,8,power_law_1.2,0.030880001187324525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,4,balanced,0.1418613294760386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,4,balanced,0.14079466462135315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,8,power_law_1.2,0.03258880078792572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,4,balanced,0.14176533619562784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,8,power_law_1.2,0.032953599095344545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,4,balanced,0.15090133746465048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,8,power_law_1.2,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,4,balanced,0.1508853336175283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,8,power_law_1.2,0.0367680013179779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,4,balanced,0.15682133038838705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,8,power_law_1.2,0.03780480027198792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,8,power_law_1.2,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,4,balanced,0.17140267292658487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,8,power_law_1.2,0.04956159889698029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,4,balanced,0.1808639963467916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,8,power_law_1.2,0.046911999583244324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,4,balanced,0.054485330979029335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,8,power_law_1.2,0.0717311978340149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,4,balanced,0.20440000295639038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,4,balanced,0.05743999779224396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,8,power_law_1.2,0.07328640222549439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,4,balanced,0.061386664708455406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,4,balanced,0.2222986618677775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,4,balanced,0.07085866729418437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,8,power_law_1.2,0.07543039917945862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,4,balanced,0.08790399630864461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,8,power_law_1.2,0.08617600202560424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,4,balanced,0.11838933825492859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,4,balanced,0.1172160009543101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.2,0.09163519740104675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,4,balanced,0.27033599217732746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,4,balanced,0.11808533469835918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.2,0.10209280252456665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,4,balanced,0.11965333422025044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.2,0.11267199516296386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,4,balanced,0.11906666556994121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,4,balanced,0.11815466483434041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.2,0.13278080224990846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,4,balanced,0.31539734204610187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,4,balanced,0.12296000123023987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.2,0.15411839485168458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,4,balanced,0.12614933649698892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.2,0.20896000862121583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,4,balanced,0.12533866365750632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.2,0.2632384061813354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,4,balanced,0.1355413297812144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,4,balanced,0.4478773276011149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.2,0.36729600429534914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,4,balanced,0.13858667016029358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,4,balanced,0.1431946655114492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.2,0.46951680183410643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,4,balanced,0.060496002435684204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,8,power_law_1.2,0.5632959842681885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,4,balanced,0.16115199526151022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,4,balanced,0.06497066716353099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,8,power_law_1.2,0.8755647659301757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,4,balanced,0.06876266499360402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,4,balanced,0.1723946730295817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,4,balanced,0.5395840009053549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,4,balanced,0.08375466863314311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,4,balanced,0.10813867052396138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,8,power_law_1.2,1.7870975494384767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,4,balanced,0.2023786703745524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,4,balanced,0.1627786656220754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,4,balanced,0.16606400410334268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,4,balanced,0.16263467073440552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,4,balanced,0.2283466657002767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,4,balanced,0.167903999487559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,4,balanced,0.15874133507410684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,4,balanced,0.15588266650835672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,4,balanced,0.2877226670583089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,4,balanced,0.7656213442484537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,4,balanced,0.1578879952430725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,4,balanced,0.159578671058019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,4,balanced,0.158160001039505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,4,balanced,0.3425866762797038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,4,balanced,0.16618133584658304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,4,balanced,0.1672746737798055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,4,balanced,0.1701386570930481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,4,balanced,0.18450667460759482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,4,balanced,0.5018080075581869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,4,balanced,0.19453867276509604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,4,balanced,0.9748373031616211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,4,balanced,0.22030399243036905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,4,balanced,0.24436799685160318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,4,balanced,0.6187359889348348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,4,balanced,0.2922133406003316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,4,balanced,0.34256001313527423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,4,balanced,0.8876586755116781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,4,balanced,1.1857279936472576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,4,balanced,0.5002613464991251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,4,balanced,1.1502292950948079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,4,balanced,0.6066453456878662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,4,balanced,1.7937547365824382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,4,balanced,0.8579733371734619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,4,balanced,1.4122506777445476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,4,balanced,1.112005313237508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,4,balanced,2.195701281229655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,4,balanced,1.3656479517618816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,4,balanced,3.5146026611328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,4,balanced,0.023402666052182514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,4,balanced,0.02480533222357432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,4,balanced,0.02629333237806956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,4,balanced,0.030245333909988403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,4,balanced,0.032730666299661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,4,balanced,0.04251199960708618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,4,balanced,0.046682665745417275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,4,balanced,0.049413333336512245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,4,balanced,0.0488373339176178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,4,balanced,2.1180639266967773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,4,balanced,0.051221330960591636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,4,balanced,0.05524266759554545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,4,balanced,0.05503466725349426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,4,balanced,0.05878399809201559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,4,balanced,0.06088533500830332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,4,balanced,0.07473599910736084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,4,balanced,0.07512533167997996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,4,balanced,4.326143900553386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,4,balanced,0.0821973333756129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,4,balanced,0.11845333377520244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,4,balanced,0.12437867124875386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,4,balanced,0.14560533563296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,4,balanced,0.16028799613316855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,4,balanced,0.1909493406613668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,4,balanced,0.21213332811991373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,4,balanced,0.2715199987093608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,4,balanced,0.30667734146118164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,4,balanced,0.4763786792755127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,4,balanced,4.155146598815918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,4,balanced,0.581061323483785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,4,balanced,0.7390986283620199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,4,power_law_1.01,0.07128959894180298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,4,power_law_1.01,0.08467199802398681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,4,power_law_1.01,0.07354239821434021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,4,power_law_1.01,0.08413439989089966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,4,power_law_1.01,0.09743360280990601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,4,power_law_1.01,0.1039423942565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,4,power_law_1.01,0.12082560062408447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,4,power_law_1.01,0.12624000310897826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,4,power_law_1.01,0.12528640031814575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,4,power_law_1.01,0.12465920448303222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,4,power_law_1.01,0.12634240388870238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,4,power_law_1.01,0.13288320302963258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,4,power_law_1.01,0.13220479488372802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,4,power_law_1.01,0.13564159870147705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,4,power_law_1.01,0.14520319700241088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,4,power_law_1.01,0.14526720046997071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,4,power_law_1.01,0.15644799470901488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,4,balanced,1.1432853539784749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,4,power_law_1.01,0.1710592031478882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,4,power_law_1.01,0.1863168001174927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,4,power_law_1.01,0.21842560768127442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,4,power_law_1.01,0.2654400110244751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,4,power_law_1.01,0.3158143997192383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,4,power_law_1.01,0.37601280212402344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,4,power_law_1.01,0.5020864009857178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,4,power_law_1.01,0.6269504070281983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,4,power_law_1.01,0.8536576271057129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,4,power_law_1.01,0.06514559984207154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,4,power_law_1.01,1.119161605834961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,4,power_law_1.01,0.07157760262489318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,4,power_law_1.01,1.3378944396972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,4,power_law_1.01,0.0644927978515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,4,power_law_1.01,2.139839935302734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,4,power_law_1.01,0.07340800166130065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,4,power_law_1.01,0.08446720242500305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,4,power_law_1.01,3.967494583129883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,4,power_law_1.01,0.09077759981155395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,4,power_law_1.01,0.10410239696502685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,4,power_law_1.01,0.10897279977798462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,4,power_law_1.01,0.10794880390167236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,4,power_law_1.01,0.1093567967414856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,4,power_law_1.01,0.11345920562744141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,4,power_law_1.01,0.1167296051979065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,4,power_law_1.01,0.12049920558929443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,4,power_law_1.01,0.12068480253219604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,4,power_law_1.01,0.12887040376663209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,4,power_law_1.01,0.1327936053276062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,4,power_law_1.01,0.14048000574111938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,4,power_law_1.01,0.15932799577713014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,4,power_law_1.01,0.17655680179595948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,4,power_law_1.01,0.21517438888549806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,4,power_law_1.01,0.2547008037567139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,4,power_law_1.01,0.3243904113769531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,4,power_law_1.01,0.391430401802063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,4,power_law_1.01,0.5500736236572266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,4,balanced,2.2467039426167807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,4,power_law_1.01,0.6972544193267822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,4,power_law_1.01,1.0065600395202636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,4,power_law_1.01,1.299180793762207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,4,power_law_1.01,1.6656576156616212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,4,power_law_1.01,2.5663103103637694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,4,power_law_1.01,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,4,power_law_1.01,5.056032180786133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,4,power_law_1.01,0.08714879751205444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,4,power_law_1.01,0.07262719869613647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,4,power_law_1.01,0.07859839797019959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,4,power_law_1.01,0.10181759595870972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,4,power_law_1.01,0.10840959548950195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,4,power_law_1.01,0.1383296012878418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,4,power_law_1.01,0.1415743947029114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,4,power_law_1.01,0.14498560428619384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,4,power_law_1.01,0.14647040367126465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,4,power_law_1.01,0.14712320566177367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,4,power_law_1.01,0.14894720315933227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,4,power_law_1.01,0.14797439575195312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,4,power_law_1.01,0.15136640071868895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,4,power_law_1.01,0.15531519651412964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,4,power_law_1.01,0.1583616018295288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,4,power_law_1.01,0.16906880140304564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,4,power_law_1.01,0.18263039588928223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.01,0.19865599870681763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.01,0.23182079792022706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.01,0.27075839042663574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.01,0.33968000411987304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.01,0.4206528186798096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.01,0.5403200149536133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.01,0.7089216232299804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.01,0.995910358428955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.01,1.304736042022705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,4,power_law_1.01,1.574796772003174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,4,power_law_1.01,2.3388160705566405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,4,power_law_1.01,4.688198471069336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,4,power_law_1.01,0.020793600380420683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,4,power_law_1.01,0.023014399409294128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,4,power_law_1.01,0.024940800666809083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,4,power_law_1.01,0.02855679988861084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,4,power_law_1.01,0.03036159873008728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,4,power_law_1.01,0.03455359935760498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,4,power_law_1.01,0.037036800384521486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,4,power_law_1.01,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,4,power_law_1.01,0.03973119854927063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,4,power_law_1.01,0.040064001083374025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,4,power_law_1.01,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,4,power_law_1.01,0.04431360065937042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,4,power_law_1.01,0.05846400260925293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,4,power_law_1.01,0.049721598625183105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,4,power_law_1.01,0.058412802219390866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,4,power_law_1.01,0.05811840295791626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,4,power_law_1.01,0.06484479904174804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,4,power_law_1.01,0.09054080247879029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.01,0.100108802318573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.01,0.11532800197601319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.01,0.12447999715805054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.01,0.156550395488739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.01,0.17842559814453124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.01,0.22483839988708496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.01,0.29018239974975585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,4,power_law_1.2,0.07061759829521179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.01,0.38689279556274414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,4,power_law_1.2,0.08146560192108154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.01,0.4758592128753662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,4,power_law_1.2,0.07288320064544677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,4,power_law_1.2,0.08714879751205444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,4,power_law_1.01,0.5981887817382813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,4,power_law_1.2,0.09218559861183166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,4,power_law_1.01,0.9353983879089356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,4,power_law_1.2,0.10165760517120362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,4,power_law_1.01,1.7102399826049806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,4,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,4,power_law_1.2,0.11885440349578857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,4,power_law_1.2,0.07033600211143494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,4,power_law_1.2,0.12451839447021484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,4,power_law_1.2,0.06220800280570984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,4,power_law_1.2,0.12202880382537842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,4,power_law_1.2,0.07564799785614014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,4,power_law_1.2,0.12992000579833984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,4,power_law_1.2,0.12721279859542847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,4,power_law_1.2,0.08009600043296813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,4,power_law_1.2,0.13141119480133057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,4,power_law_1.2,0.08814719915390015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,4,power_law_1.2,0.13492480516433716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,4,power_law_1.2,0.10410239696502685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,4,power_law_1.2,0.13642879724502563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,4,power_law_1.2,0.11055999994277954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,4,power_law_1.2,0.1454848051071167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,4,power_law_1.2,0.10659199953079224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,4,power_law_1.2,0.14940160512924194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,4,power_law_1.2,0.10880639553070068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,4,power_law_1.2,0.16131839752197266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,4,power_law_1.2,0.11275520324707031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,4,power_law_1.2,0.1763584017753601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,4,power_law_1.2,0.11587200164794922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,4,power_law_1.2,0.19187840223312377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,4,power_law_1.2,0.11765120029449463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,4,power_law_1.2,0.23758080005645751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,4,power_law_1.2,0.12390400171279907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,4,power_law_1.2,0.2670783996582031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,4,power_law_1.2,0.1312000036239624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,4,power_law_1.2,0.33006720542907714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,4,power_law_1.2,0.13494399785995484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,4,power_law_1.2,0.14350719451904298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,4,power_law_1.2,0.40446720123291013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,4,power_law_1.2,0.16906239986419677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,4,power_law_1.2,0.5342720031738282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,4,power_law_1.2,0.18101760149002075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,4,power_law_1.2,0.6949120044708252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,4,power_law_1.2,0.2291264057159424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,4,power_law_1.2,0.9518976211547852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,4,power_law_1.2,0.26525440216064455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,4,power_law_1.2,1.1976767539978028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,4,power_law_1.2,0.34088959693908694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,4,power_law_1.2,1.3920703887939454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,4,power_law_1.2,0.4119743824005127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,4,power_law_1.2,2.513868713378906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,4,power_law_1.2,0.599398422241211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,4,power_law_1.2,4.565254211425781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,4,power_law_1.2,0.7488319873809814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,4,power_law_1.2,1.0257856369018554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,4,power_law_1.2,1.3698111534118653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,4,power_law_1.2,1.6044672012329102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,4,power_law_1.2,2.494009590148926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,4,power_law_1.2,0.0732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,4,power_law_1.2,0.08171520233154297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,4,power_law_1.2,5.206911849975586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,4,power_law_1.2,0.06947839856147767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,4,power_law_1.2,0.08092799782752991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,4,power_law_1.2,0.09473279714584351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,4,power_law_1.2,0.10543359518051147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,4,power_law_1.2,0.13329919576644897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,4,power_law_1.2,0.1364799976348877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,4,power_law_1.2,0.14290560483932496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,4,power_law_1.2,0.13875839710235596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,4,power_law_1.2,0.14165120124816893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,4,power_law_1.2,0.146943998336792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,4,power_law_1.2,0.14847359657287598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,4,power_law_1.2,0.1572991967201233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,4,power_law_1.2,0.16053119897842408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,4,power_law_1.2,0.16165119409561157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,4,power_law_1.2,0.1718016028404236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,4,power_law_1.2,0.18855680227279664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.2,0.20979199409484864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.2,0.24096639156341554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.2,0.27720959186553956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.2,0.3618240118026733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.2,0.4336512088775635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.2,0.580787181854248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.2,0.7667200088500976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.2,1.094700813293457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.2,1.3125120162963868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,4,power_law_1.2,1.6340927124023437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,4,power_law_1.2,2.5576831817626955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,4,power_law_1.2,5.355807876586914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,4,power_law_1.2,0.02067199945449829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,4,power_law_1.2,0.02268799990415573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,4,power_law_1.2,0.02335360050201416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,4,power_law_1.2,0.028230398893356323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,4,power_law_1.2,0.029875200986862183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,4,power_law_1.2,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,4,power_law_1.2,0.036934399604797365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,4,power_law_1.2,0.04014720022678375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,4,power_law_1.2,0.0387584000825882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,4,power_law_1.2,0.04044159948825836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,4,power_law_1.2,0.04380159974098206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,4,power_law_1.2,0.045075199007987975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,4,power_law_1.2,0.04576640129089356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,4,power_law_1.2,0.04876160025596619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,4,power_law_1.2,0.058208000659942624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,4,power_law_1.2,0.061926400661468504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,4,power_law_1.2,0.06778879761695862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,4,power_law_1.2,0.0914687991142273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.2,0.11746560335159302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.2,0.12223360538482667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.2,0.13351680040359498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.2,0.15649919509887694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.2,0.18045439720153808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.2,0.2332927942276001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.2,0.2838144063949585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.2,0.3837183952331543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.2,0.5077055931091309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,2,balanced,0.06657066444555919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,4,power_law_1.2,0.6030911922454834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,2,balanced,0.06241066753864288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,2,balanced,0.07042666773001353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,2,balanced,0.06356800099213918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,2,balanced,0.07700799902280171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,4,power_law_1.2,0.9324799537658691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,2,balanced,0.07170666754245758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,2,balanced,0.09708799918492635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,2,balanced,0.08639466762542725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,2,balanced,0.12567999958992004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,4,power_law_1.2,1.8098175048828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,2,balanced,0.11076266566912334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,2,balanced,0.17009600003560385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,2,balanced,0.1597813367843628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,2,balanced,0.17171732584635416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,2,balanced,0.15852266550064087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,2,balanced,0.1707520087560018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,2,balanced,0.15967999895413718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,2,balanced,0.17216000954310098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,2,balanced,0.1633386711279551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,2,balanced,0.1728586753209432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,2,balanced,0.16423466801643372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,2,balanced,0.1743839979171753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,2,balanced,0.16378666957219443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,2,balanced,0.17910399039586386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,2,balanced,0.1662613352139791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,2,balanced,0.17853333552678427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,2,balanced,0.169978658358256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,2,balanced,0.17203199863433838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,2,balanced,0.17931199073791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,2,balanced,0.18189332882563272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,2,balanced,0.19021866718928018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,2,balanced,0.18476800123850504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,2,balanced,0.1922826568285624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,2,balanced,0.1914773384730021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,2,balanced,0.20094933112462363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,2,balanced,0.21626132726669312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,2,balanced,0.22181334098180136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,2,balanced,0.2336853345235189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,2,balanced,0.23563732703526816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,2,balanced,0.27793065706888836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,2,balanced,0.2727573315302531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,2,balanced,0.31174933910369873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,2,balanced,0.30372800429662067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,2,balanced,0.40139734745025635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,2,balanced,0.37828266620635986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,2,balanced,0.4827626546223958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,2,balanced,0.45098666350046795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,2,balanced,0.7261866728464762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,2,balanced,0.6586079994837443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,2,balanced,0.8959253629048666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,2,power_law_1.01,0.07054719924926758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,2,power_law_1.01,0.07834879755973816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,2,power_law_1.01,0.07768319845199585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,2,balanced,0.8046720027923584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,2,power_law_1.01,0.09244800209999085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,2,power_law_1.01,0.10794880390167236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,2,balanced,1.300821304321289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,2,power_law_1.01,0.12312320470809937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,2,power_law_1.01,0.14058879613876343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,2,power_law_1.01,0.14744319915771484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,2,power_law_1.01,0.14981119632720946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,2,power_law_1.01,0.15114879608154297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,2,power_law_1.01,0.15799039602279663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,2,power_law_1.01,0.16202880144119264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,2,power_law_1.01,0.1640064001083374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,2,balanced,1.1484533150990803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,2,power_law_1.01,0.17031680345535277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,2,balanced,1.6949440638224285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,2,power_law_1.01,0.1788032054901123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,2,power_law_1.01,0.1819584012031555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,2,power_law_1.01,0.1932479977607727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,2,power_law_1.01,0.21400320529937744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,2,power_law_1.01,0.2386240005493164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,2,power_law_1.01,0.2850048065185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,2,power_law_1.01,0.3377279996871948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,2,power_law_1.01,0.4381247997283936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,2,power_law_1.01,0.5057024002075196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,2,power_law_1.01,0.7086016178131104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,2,balanced,1.4847359657287598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,2,power_law_1.01,0.8661567687988281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,2,balanced,2.0841387112935386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,2,power_law_1.01,1.2691264152526855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,2,power_law_1.01,1.6339456558227539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,2,power_law_1.01,2.0140287399291994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,2,power_law_1.01,3.040083122253418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,2,power_law_1.01,5.7512256622314455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,2,balanced,1.8196853001912434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,2,balanced,3.2571681340535483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,2,power_law_1.01,0.06522240042686463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,2,power_law_1.01,0.0725055992603302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,2,power_law_1.01,0.0671231985092163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,2,power_law_1.01,0.08707200288772583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,2,power_law_1.01,0.09968640208244324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,2,power_law_1.01,0.11163519620895386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,2,power_law_1.01,0.12878719568252564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,2,power_law_1.01,0.13927680253982544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,2,power_law_1.01,0.14021120071411133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,2,power_law_1.01,0.14206080436706542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,2,power_law_1.01,0.14695039987564087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,2,power_law_1.01,0.15179519653320311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,2,power_law_1.01,0.1542080044746399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,2,power_law_1.01,0.1615231990814209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,2,balanced,2.8058239618937173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,2,power_law_1.01,0.17228800058364868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,2,power_law_1.01,0.17797759771347046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,2,power_law_1.01,0.1857792019844055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,2,power_law_1.01,0.21512320041656494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,2,power_law_1.01,0.2425856113433838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,2,power_law_1.01,0.290118408203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,2,power_law_1.01,0.3496000051498413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,2,power_law_1.01,0.45548157691955565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,2,power_law_1.01,0.5541247844696044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,2,power_law_1.01,0.7703487873077393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,2,balanced,6.436682383219401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,2,power_law_1.01,0.9844672203063964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,2,power_law_1.01,1.3884991645812987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,2,power_law_1.01,1.8392831802368164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,2,power_law_1.01,2.199910354614258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,2,power_law_1.01,3.523020935058594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,2,power_law_1.01,7.099910736083984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,2,balanced,5.513546625773112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,2,power_law_1.2,0.06993920207023621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,2,power_law_1.2,0.07734400033950806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,2,power_law_1.2,0.07921919822692872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,2,power_law_1.2,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,2,power_law_1.2,0.10467840433120727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,2,power_law_1.2,0.11916799545288086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,2,power_law_1.2,0.139628803730011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,2,power_law_1.2,0.14345599412918092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,2,power_law_1.2,0.1493824005126953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,2,power_law_1.2,0.1477952003479004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,2,power_law_1.2,0.15437439680099488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,2,power_law_1.2,0.16005120277404786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,2,power_law_1.2,0.16373759508132935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,2,power_law_1.2,0.17169920206069947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,2,power_law_1.2,0.1826815962791443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,2,power_law_1.2,0.18719359636306762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,2,power_law_1.2,0.19881600141525269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,2,power_law_1.2,0.21898880004882812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,2,power_law_1.2,0.24527359008789062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,2,power_law_1.2,0.29875199794769286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,2,power_law_1.2,0.3533375978469849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,2,power_law_1.2,0.433625602722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,2,power_law_1.2,0.5161280155181884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,2,power_law_1.2,0.7229184150695801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,2,power_law_1.2,0.9332927703857422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,2,power_law_1.2,1.2212160110473633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,2,power_law_1.2,1.643974494934082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,2,power_law_1.2,2.002239990234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,2,power_law_1.2,3.2293632507324217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,2,power_law_1.2,0.06538239717483521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,2,power_law_1.2,6.74316177368164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,2,power_law_1.2,0.07271040081977845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,2,power_law_1.2,0.06892160177230836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,2,power_law_1.2,0.08684160113334656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,2,power_law_1.2,0.0969215989112854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,2,power_law_1.2,0.11331199407577515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,64,balanced,0.03405333310365677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,2,power_law_1.2,0.1300287961959839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,64,balanced,0.03499199946721395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,2,power_law_1.2,0.13600640296936034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,64,balanced,0.03401600072781245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,64,balanced,0.035071998834609985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,2,power_law_1.2,0.14435839653015137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,64,balanced,0.039146666725476585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,2,power_law_1.2,0.14309120178222656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,64,balanced,0.036720000207424164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,2,power_law_1.2,0.14326399564743042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,64,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,64,balanced,0.042064001162846885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,2,power_law_1.2,0.1527168035507202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,64,balanced,0.03876800090074539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,2,power_law_1.2,0.1548416018486023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,64,balanced,0.03835733234882355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,64,balanced,0.03849600007136663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,2,power_law_1.2,0.16273280382156372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,64,balanced,0.03914133210976919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,2,power_law_1.2,0.1719488024711609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,64,balanced,0.04012266546487808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,2,power_law_1.2,0.1793984055519104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,64,balanced,0.0397173340121905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,2,power_law_1.2,0.1932479977607727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,64,balanced,0.04246933261553446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,2,power_law_1.2,0.21913599967956543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,64,balanced,0.04442666471004486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,2,power_law_1.2,0.24318079948425292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,64,balanced,0.04409066836039225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,2,power_law_1.2,0.29173760414123534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,2,power_law_1.2,0.3546432018280029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,64,balanced,0.044693330923716225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,2,power_law_1.2,0.45574398040771485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,64,balanced,0.043978666265805565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,2,power_law_1.2,0.5793983936309814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,2,power_law_1.2,0.7763775825500489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,64,balanced,0.04677333434422811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,2,power_law_1.2,0.9617216110229492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,64,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,2,power_law_1.2,1.4438207626342774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,2,power_law_1.2,1.8714176177978517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,64,balanced,0.05014933149019877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,2,power_law_1.2,2.162144088745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,2,power_law_1.2,3.685408020019531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,64,balanced,0.053818667928377785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,2,power_law_1.2,7.072621154785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,64,balanced,0.032986665765444435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,64,balanced,0.034661332766215004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,64,balanced,0.06221333146095276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,64,balanced,0.033514666060606636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,64,balanced,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,64,balanced,0.03882666677236557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,64,balanced,0.038912000755469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,64,balanced,0.03689600030581156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,64,balanced,0.03912533322970072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,64,balanced,0.07044800122578938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,64,balanced,0.038693333665529885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,64,balanced,0.03839466720819473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,64,balanced,0.03904533386230469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,64,balanced,0.03841066608826319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,64,balanced,0.04058133314053217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,64,balanced,0.04065600037574768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,64,balanced,0.04275733232498169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,64,balanced,0.08903466661771138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,64,balanced,0.04500266909599304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,64,balanced,0.04470933477083842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,64,balanced,0.046629334489504494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,64,balanced,0.046485334634780884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,64,balanced,0.050704002380371094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,64,balanced,0.05282666782538096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,64,balanced,0.11102400223414104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,64,balanced,0.05917333563168844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,64,balanced,0.06256533165772755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,64,balanced,0.08187200129032135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,64,balanced,0.12637866536776224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,64,balanced,0.10809066891670227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,64,balanced,0.034058667719364166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,64,balanced,0.035088000198205314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,64,balanced,0.14656000336011252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,64,balanced,0.03465600063403448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,64,balanced,0.036677333215872444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,64,balanced,0.03909866760174433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,64,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,64,balanced,0.040106666584809623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,64,balanced,0.1809013287226359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,64,balanced,0.040565334260463715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,64,balanced,0.038506666819254555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,64,balanced,0.040192000567913055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,64,balanced,0.04026666780312856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,64,balanced,0.041077333192030586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,64,balanced,0.03854399919509888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,64,balanced,0.04090133309364319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,64,balanced,0.044682666659355164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,64,balanced,0.18093866109848022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,64,balanced,0.04444799820582072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,64,balanced,0.044533332188924156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,64,balanced,0.04486933350563049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,64,balanced,0.048341333866119385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,64,balanced,0.04821333289146423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,64,balanced,0.0528053343296051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,64,balanced,0.05629866818586985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,64,balanced,0.2156053384145101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,64,balanced,0.06253333389759064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,64,balanced,0.07595733304818471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,64,balanced,0.01950399950146675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,64,balanced,0.02019199977318446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,64,balanced,0.01974933346112569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,64,balanced,0.020015999674797058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,64,balanced,0.08917333682378133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,64,balanced,0.0223786657055219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,64,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,64,balanced,0.030666666726271313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,64,balanced,0.03010133405526479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,64,balanced,0.030181333422660828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,64,balanced,0.030069333811601002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,64,balanced,0.030586667358875275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,64,balanced,0.319269339243571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,64,balanced,0.0296426663796107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,64,balanced,0.02979733298222224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,64,balanced,0.11426132917404175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,64,balanced,0.03028800090154012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,64,balanced,0.31651200850804645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,64,balanced,0.03426666557788849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,64,balanced,0.032629333436489105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,64,balanced,0.03303466737270355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,64,balanced,0.035071998834609985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,64,balanced,0.036757332583268486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,64,balanced,0.04062933226426443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,64,balanced,0.13863999644915262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,64,balanced,0.050661335388819374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,64,balanced,0.056559999783833824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,64,balanced,0.06690133114655812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,64,balanced,0.1627786656220754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,64,balanced,0.09114666779836018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,64,balanced,0.11381866534550984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,64,balanced,0.15689599514007568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,64,balanced,0.5901600122451782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,64,balanced,0.23062400023142496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,64,balanced,0.1999839941660563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,64,balanced,0.2419253389040629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,64,balanced,0.4187839825948079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,64,power_law_1.01,0.03415040075778961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,64,power_law_1.01,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,64,power_law_1.01,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,64,power_law_1.01,0.036032000184059144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,64,power_law_1.01,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,64,power_law_1.01,0.03826560080051422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,64,balanced,0.3704853455225627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,64,power_law_1.01,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,64,power_law_1.01,0.03699199855327606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,64,power_law_1.01,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,64,power_law_1.01,0.03738879859447479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,64,power_law_1.01,0.03802880048751831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,64,power_law_1.01,0.03829759955406189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,64,power_law_1.01,0.03809280097484589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,64,power_law_1.01,0.03982079923152924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,64,power_law_1.01,0.042105600237846375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,64,power_law_1.01,0.04196479916572571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,64,power_law_1.01,0.04299519956111908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,64,power_law_1.01,0.04343039989471435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,64,power_law_1.01,0.04517120122909546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,64,power_law_1.01,0.0535103976726532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,64,power_law_1.01,0.05868160128593445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,64,power_law_1.01,0.07217919826507568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,64,power_law_1.01,0.07895680069923401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,64,power_law_1.01,0.108787202835083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,64,power_law_1.01,0.12983039617538453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,64,power_law_1.01,0.15463039875030518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,64,power_law_1.01,0.23172481060028077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,64,power_law_1.01,0.2425920009613037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,64,power_law_1.01,0.4253568172454834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,64,power_law_1.01,0.8210047721862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,64,power_law_1.01,0.03407360017299652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,64,balanced,0.7150506973266602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,64,power_law_1.01,0.0339711993932724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,64,power_law_1.01,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,64,power_law_1.01,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,64,power_law_1.01,0.0379967987537384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,64,power_law_1.01,0.03671039938926697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,64,power_law_1.01,0.037190398573875426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,64,power_law_1.01,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,64,power_law_1.01,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,64,power_law_1.01,0.03743360042572021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,64,power_law_1.01,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,64,power_law_1.01,0.038815999031066896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,64,power_law_1.01,0.03891200125217438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,64,power_law_1.01,0.038924801349639895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,64,power_law_1.01,0.042745599150657655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,64,power_law_1.01,0.04311679899692535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,64,power_law_1.01,0.04357759952545166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,64,power_law_1.01,0.04792959988117218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,64,power_law_1.01,0.05023360252380371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,64,power_law_1.01,0.05618559718132019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,64,power_law_1.01,0.060166400671005246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,64,power_law_1.01,0.07516800165176392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,64,power_law_1.01,0.08845440149307252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,64,power_law_1.01,0.12271360158920289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,64,power_law_1.01,0.15797120332717896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,64,power_law_1.01,0.22588160037994384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,64,power_law_1.01,0.2899456024169922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,64,power_law_1.01,0.357043194770813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,64,power_law_1.01,0.5536640167236329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,64,power_law_1.01,0.03389439880847931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,64,power_law_1.01,1.0805184364318847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,64,power_law_1.01,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,64,power_law_1.01,0.03436799943447113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,64,power_law_1.01,0.03656319975852966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,64,power_law_1.01,0.03691520094871521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,64,power_law_1.01,0.03841919898986816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,64,power_law_1.01,0.03795199990272522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,64,power_law_1.01,0.037920001149177554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,64,power_law_1.01,0.037964800000190736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,64,power_law_1.01,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,64,power_law_1.01,0.03839359879493713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,64,power_law_1.01,0.03870719969272614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,64,power_law_1.01,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,64,power_law_1.01,0.03935360014438629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,64,power_law_1.01,0.041657599806785586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,64,power_law_1.01,0.04257920086383819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,64,power_law_1.01,0.04407039880752563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,64,power_law_1.01,0.04912000000476837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,64,power_law_1.01,0.05067520141601563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,64,power_law_1.01,0.05754240155220032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,64,power_law_1.01,0.06495360136032105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,64,power_law_1.01,0.06973440051078797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,64,power_law_1.01,0.0818560004234314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,64,power_law_1.01,0.1097599983215332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,64,power_law_1.01,0.13191039562225343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,64,power_law_1.01,0.1866495966911316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,64,power_law_1.01,0.23683199882507325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,64,power_law_1.01,0.2572160005569458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,64,power_law_1.01,0.5201856136322022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,64,power_law_1.01,0.9087679862976075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,64,power_law_1.01,0.01741439998149872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,64,power_law_1.01,0.018156799674034118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,64,power_law_1.01,0.01794559955596924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,64,power_law_1.01,0.018598400056362152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,64,power_law_1.01,0.020633600652217865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,64,power_law_1.01,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,64,power_law_1.01,0.029049599170684816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,64,power_law_1.01,0.02807680070400238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,64,power_law_1.01,0.028300800919532777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,64,power_law_1.01,0.028537601232528687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,64,power_law_1.01,0.028825598955154418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,64,power_law_1.01,0.02901119887828827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,64,power_law_1.01,0.028755199909210206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,64,power_law_1.01,0.029452800750732422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,64,power_law_1.01,0.03022719919681549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,64,power_law_1.01,0.030988800525665283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,64,power_law_1.01,0.03237760066986084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,64,power_law_1.2,0.0353983998298645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,64,power_law_1.01,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,64,power_law_1.2,0.03603839874267578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,64,power_law_1.01,0.035872000455856326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,64,power_law_1.2,0.03425920009613037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,64,power_law_1.01,0.039526399970054624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,64,power_law_1.2,0.03682560026645661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,64,power_law_1.01,0.04618239998817444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,64,power_law_1.2,0.037299200892448425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,64,power_law_1.01,0.054079997539520266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,64,power_law_1.2,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,64,power_law_1.01,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,64,power_law_1.2,0.03722879886627197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,64,power_law_1.01,0.08151040077209473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,64,power_law_1.2,0.038540801405906676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,64,power_law_1.01,0.1003648042678833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,64,power_law_1.2,0.0367935985326767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,64,power_law_1.01,0.15089279413223267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,64,power_law_1.2,0.037241598963737486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,64,power_law_1.2,0.03835520148277283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,64,power_law_1.01,0.17744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,64,power_law_1.2,0.03850240111351013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,64,power_law_1.01,0.21381759643554688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,64,power_law_1.2,0.03930239975452423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,64,power_law_1.01,0.3338752031326294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,64,power_law_1.2,0.03951359987258911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,64,power_law_1.2,0.04215039908885956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,64,power_law_1.01,0.629311990737915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,64,power_law_1.2,0.042559999227523806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,64,power_law_1.2,0.04366079866886139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,64,power_law_1.2,0.04709759950637817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,64,power_law_1.2,0.0496832013130188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,64,power_law_1.2,0.05725439786911011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,64,power_law_1.2,0.06913920044898987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,64,power_law_1.2,0.0781440019607544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,64,power_law_1.2,0.09393919706344604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,64,power_law_1.2,0.14535679817199706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,64,power_law_1.2,0.1733888030052185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,64,power_law_1.2,0.23503360748291016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,64,power_law_1.2,0.3586047887802124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,64,power_law_1.2,0.47727360725402834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,64,power_law_1.2,0.6901951789855957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,64,power_law_1.2,1.580454444885254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,64,power_law_1.2,0.034227201342582704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,64,power_law_1.2,0.03276160061359405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,64,power_law_1.2,0.03251200020313263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,64,power_law_1.2,0.035392001271247864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,64,power_law_1.2,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,64,power_law_1.2,0.0364544004201889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,64,power_law_1.2,0.03680639863014221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,64,power_law_1.2,0.036287999153137206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,64,power_law_1.2,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,64,power_law_1.2,0.03704319894313812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,64,power_law_1.2,0.03784320056438446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,64,power_law_1.2,0.03782399892807007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,64,power_law_1.2,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,64,power_law_1.2,0.03984000086784363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,64,power_law_1.2,0.04216319918632507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,64,power_law_1.2,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,64,power_law_1.2,0.04376960098743439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,64,power_law_1.2,0.04813440144062042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,64,power_law_1.2,0.05060480237007141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,64,power_law_1.2,0.06356480121612548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,64,power_law_1.2,0.06858879923820496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,64,power_law_1.2,0.08915200233459472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,64,power_law_1.2,0.1109503984451294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,64,power_law_1.2,0.15463039875030518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,64,power_law_1.2,0.21750400066375733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,64,power_law_1.2,0.2821055889129639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,64,power_law_1.2,0.38918399810791016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,64,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,64,power_law_1.2,0.4949312210083008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,64,power_law_1.2,0.03386240005493164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,64,power_law_1.2,0.03255679905414581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,64,power_law_1.2,0.8707648277282715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,64,power_law_1.2,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,64,power_law_1.2,2.002899169921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,64,power_law_1.2,0.03711999952793121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,64,power_law_1.2,0.037503999471664426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,64,power_law_1.2,0.03764480054378509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,64,power_law_1.2,0.038540801405906676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,64,power_law_1.2,0.037574398517608645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,64,power_law_1.2,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,64,power_law_1.2,0.03860479891300202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,64,power_law_1.2,0.03894400000572205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,64,power_law_1.2,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,64,power_law_1.2,0.040217599272727965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,64,power_law_1.2,0.043244799971580504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,64,power_law_1.2,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,64,power_law_1.2,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,64,power_law_1.2,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,64,power_law_1.2,0.05292159914970398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,64,power_law_1.2,0.05867519974708557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,64,power_law_1.2,0.06709120273590088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,64,power_law_1.2,0.08310400247573853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,64,power_law_1.2,0.09071360230445862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,64,power_law_1.2,0.14264320135116576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,64,power_law_1.2,0.1843840003013611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,64,power_law_1.2,0.23690240383148192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,64,power_law_1.2,0.333516788482666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,64,power_law_1.2,0.4272768020629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,64,power_law_1.2,0.6671872138977051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,64,power_law_1.2,1.5269696235656738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,64,power_law_1.2,0.01727360039949417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,64,power_law_1.2,0.016991999745368958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,64,power_law_1.2,0.01849599927663803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,64,power_law_1.2,0.019308799505233766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,64,power_law_1.2,0.021510399878025055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,64,power_law_1.2,0.03167999982833862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,64,power_law_1.2,0.02959359884262085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,64,power_law_1.2,0.02869119942188263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,64,power_law_1.2,0.028428798913955687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,64,power_law_1.2,0.02895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,64,power_law_1.2,0.02905600070953369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,64,power_law_1.2,0.02868480086326599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,64,power_law_1.2,0.02917119860649109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,64,power_law_1.2,0.029292801022529603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,64,power_law_1.2,0.029683199524879456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,64,power_law_1.2,0.031692799925804135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,64,power_law_1.2,0.032627201080322264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,32,balanced,0.03172266731659571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,32,balanced,0.032586666444937386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,64,power_law_1.2,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,32,balanced,0.032826667030652366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,64,power_law_1.2,0.036735999584198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,32,balanced,0.036864000062147774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,64,power_law_1.2,0.04136320054531097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,32,balanced,0.03902400036652883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,32,balanced,0.03834133346875509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,64,power_law_1.2,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,32,balanced,0.03860799968242645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,64,power_law_1.2,0.053939199447631835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,32,balanced,0.038831998904546104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,64,power_law_1.2,0.06407039761543273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,32,balanced,0.0402399996916453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,64,power_law_1.2,0.08812159895896912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,32,balanced,0.03860799968242645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,64,power_law_1.2,0.11058559417724609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,32,balanced,0.03825599948565165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,64,power_law_1.2,0.14988800287246704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,32,balanced,0.040362666050593056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,32,balanced,0.038405333956082664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,64,power_law_1.2,0.19338879585266114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,32,balanced,0.040821333726247154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,64,power_law_1.2,0.2390144109725952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,32,balanced,0.04277333120505015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,64,power_law_1.2,0.3658560037612915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,32,balanced,0.042303999265034996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,64,power_law_1.2,0.6943871974945068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,32,balanced,0.044490665197372437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,32,balanced,0.044677332043647766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,32,balanced,0.04621866842110952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,32,balanced,0.04633066554864248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,32,balanced,0.048122664292653404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,32,balanced,0.05425066749254862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,32,balanced,0.05482666691144308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,32,balanced,0.06658133367697398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,32,balanced,0.03161599983771642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,32,balanced,0.03312533348798752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,32,balanced,0.03268266717592875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,32,balanced,0.07705066601435344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,32,balanced,0.03446933378775915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,32,balanced,0.039018665750821434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,32,balanced,0.03852800031503042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,32,balanced,0.03692800054947535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,32,balanced,0.03919466584920883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,32,balanced,0.03896533449490865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,32,balanced,0.03937066594759623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,32,balanced,0.038922667503356934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,32,balanced,0.038762666285037994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,32,balanced,0.09999466935793559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,32,balanced,0.04072533299525579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,32,balanced,0.04107200105985006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,32,balanced,0.042965332667032875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,32,balanced,0.04466133316357931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,32,balanced,0.044112001856168113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,32,balanced,0.0465280016263326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,32,balanced,0.04690133531888326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,32,balanced,0.0525546669960022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,32,balanced,0.12432000041007996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,32,balanced,0.05241066714127859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,32,balanced,0.06049066781997681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,32,balanced,0.06596800188223521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,32,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,32,balanced,0.03438399980465571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,32,balanced,0.1449120044708252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,32,balanced,0.03456533451875051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,32,balanced,0.0892639954884847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,32,balanced,0.036650667587916054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,32,balanced,0.04088533421357473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,32,balanced,0.04015466570854187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,32,balanced,0.03867733230193456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,32,balanced,0.04091199984153112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,32,balanced,0.040618665516376495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,32,balanced,0.04092800120512644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,32,balanced,0.1183093289534251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,32,balanced,0.04028266668319702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,32,balanced,0.04067199925581614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,32,balanced,0.04301333427429199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,32,balanced,0.04243200023969015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,32,balanced,0.0443146675825119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,32,balanced,0.04620266457398733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,32,balanced,0.046709333856900535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,32,balanced,0.04808533191680908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,32,balanced,0.15847466389338175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,32,balanced,0.048341333866119385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,32,balanced,0.05282133320967356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,32,balanced,0.2020053267478943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,32,balanced,0.054842665791511536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,32,balanced,0.0627040018637975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,32,balanced,0.19964800278345743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,32,balanced,0.06709333260854085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,32,balanced,0.01757866640885671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,32,balanced,0.019648000597953796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,32,balanced,0.019744000087181728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,32,balanced,0.021189334491888683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,32,balanced,0.020090666910012562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,32,balanced,0.022895999252796173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,32,balanced,0.08562133709589641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,32,balanced,0.034373333056767784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,32,balanced,0.03417066733042399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,32,balanced,0.029887999097506206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,32,balanced,0.030479999879995983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,32,balanced,0.030224000414212544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,32,balanced,0.02997333308060964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,32,balanced,0.23573333024978638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,32,balanced,0.029626667499542236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,32,balanced,0.030016000072161358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,32,balanced,0.10214400291442871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,32,balanced,0.0323840007185936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,32,balanced,0.032730666299661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,32,balanced,0.03250666707754135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,32,balanced,0.03618133316437403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,32,balanced,0.03973866750796636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,32,balanced,0.044533332188924156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,32,balanced,0.05175999800364176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,32,balanced,0.1306613286336263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,32,balanced,0.06239999830722809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,32,balanced,0.3678079843521118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,32,balanced,0.3502986828486125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,32,balanced,0.07666666805744171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,32,balanced,0.10132267077763875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,32,balanced,0.1593706707159678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,32,balanced,0.12826666235923767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,32,balanced,0.17453332742055258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,32,balanced,0.18785067399342856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,32,balanced,0.2222399910291036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,32,balanced,0.6524106661478678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,32,balanced,0.26950399080912274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,32,balanced,0.27190399169921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,32,balanced,0.4157493511835734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,32,power_law_1.01,0.03566080033779144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,32,power_law_1.01,0.0328575998544693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,32,power_law_1.01,0.03253760039806366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,32,balanced,0.49933866659800213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,32,power_law_1.01,0.03501439988613129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,32,power_law_1.01,0.03680639863014221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,32,power_law_1.01,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,32,power_law_1.01,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,32,balanced,0.8050346374511719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,32,power_law_1.01,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,32,power_law_1.01,0.03768320083618164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,32,power_law_1.01,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,32,power_law_1.01,0.038771200180053714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,32,power_law_1.01,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,32,power_law_1.01,0.038822400569915774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,32,power_law_1.01,0.03847039937973022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,32,power_law_1.01,0.04165120124816894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,32,power_law_1.01,0.04258559942245484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,32,power_law_1.01,0.04328320026397705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,32,power_law_1.01,0.04576640129089356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,32,power_law_1.01,0.04695039987564087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,32,power_law_1.01,0.0531711995601654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,32,power_law_1.01,0.06120319962501526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,32,power_law_1.01,0.07210239768028259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,32,power_law_1.01,0.08076800107955932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,32,power_law_1.01,0.10404479503631592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,32,power_law_1.01,0.13372160196304322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,32,power_law_1.01,0.18315520286560058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,32,power_law_1.01,0.22383999824523926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,32,power_law_1.01,0.2437376022338867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,32,power_law_1.01,0.41845121383666994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,32,power_law_1.01,0.7909567832946778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,32,power_law_1.01,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,32,power_law_1.01,0.0313728004693985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,32,power_law_1.01,0.03182080090045929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,32,power_law_1.01,0.03550080060958862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,32,power_law_1.01,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,32,power_law_1.01,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,32,power_law_1.01,0.036524799466133115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,32,power_law_1.01,0.03691520094871521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,32,power_law_1.01,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,32,power_law_1.01,0.03797119855880737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,32,power_law_1.01,0.03768959939479828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,32,power_law_1.01,0.03828479945659637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,32,power_law_1.01,0.03880960047245026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,32,power_law_1.01,0.04108160138130188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,32,power_law_1.01,0.04335359930992126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,32,power_law_1.01,0.04373759925365448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,32,power_law_1.01,0.04430719912052154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,32,power_law_1.01,0.047705599665641786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,32,power_law_1.01,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,32,power_law_1.01,0.058208000659942624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,32,power_law_1.01,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,32,power_law_1.01,0.07976319789886474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,32,power_law_1.01,0.09281920194625855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,32,power_law_1.01,0.13352960348129272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,32,power_law_1.01,0.16524159908294678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,32,power_law_1.01,0.23349759578704835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,32,power_law_1.01,0.28462719917297363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,32,power_law_1.01,0.3788032054901123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,32,power_law_1.01,0.545139217376709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,32,power_law_1.01,1.193779182434082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,32,power_law_1.01,0.03513599932193756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,32,power_law_1.01,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,32,power_law_1.01,0.03293440043926239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,32,power_law_1.01,0.0356799989938736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,32,power_law_1.01,0.03661440014839172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,32,power_law_1.01,0.016531200706958772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,32,power_law_1.01,0.03739520013332367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,32,power_law_1.01,0.01785600036382675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,32,power_law_1.01,0.03856639862060547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,32,power_law_1.01,0.017510400712490083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,32,power_law_1.01,0.039110401272773744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,32,power_law_1.01,0.017612800002098083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,32,power_law_1.01,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,32,power_law_1.01,0.018739199638366698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,32,power_law_1.01,0.03861759901046753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,32,power_law_1.01,0.021081599593162536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,32,power_law_1.01,0.03934719860553741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,32,power_law_1.01,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,32,power_law_1.01,0.04044159948825836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,32,power_law_1.01,0.03271040022373199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,32,power_law_1.01,0.02842240035533905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,32,power_law_1.01,0.03919999897480011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,32,power_law_1.01,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,32,power_law_1.01,0.028908801078796387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,32,power_law_1.01,0.04348799884319306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,32,power_law_1.01,0.028748801350593566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,32,power_law_1.01,0.0444927990436554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,32,power_law_1.01,0.02900480031967163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,32,power_law_1.01,0.028825598955154418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,32,power_law_1.01,0.04555520117282867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,32,power_law_1.01,0.02813439965248108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,32,power_law_1.01,0.05047680139541626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,32,power_law_1.01,0.02913280129432678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,32,power_law_1.01,0.05400959849357605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,32,power_law_1.01,0.030963200330734252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,32,power_law_1.01,0.058278399705886844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,32,power_law_1.01,0.03091199994087219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,32,power_law_1.01,0.06736000180244446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,32,power_law_1.01,0.03335680067539215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,32,power_law_1.01,0.0781823992729187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,32,power_law_1.01,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,32,power_law_1.01,0.09451519846916198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,32,power_law_1.01,0.04035840034484863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,32,power_law_1.01,0.11941759586334229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,32,power_law_1.01,0.04161919951438904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,32,power_law_1.01,0.1599295973777771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,32,power_law_1.01,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,32,power_law_1.01,0.19705599546432495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,32,power_law_1.2,0.03473280072212219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,32,power_law_1.01,0.06090239882469177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,32,power_law_1.2,0.03269760012626648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,32,power_law_1.01,0.24535679817199707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,32,power_law_1.01,0.07685120105743408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,32,power_law_1.2,0.032025599479675294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,32,power_law_1.01,0.30592639446258546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,32,power_law_1.01,0.09299839735031128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,32,power_law_1.2,0.03488639891147614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,32,power_law_1.01,0.5023808002471923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,32,power_law_1.2,0.03594239950180054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,32,power_law_1.01,0.1377087950706482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,32,power_law_1.01,0.9195391654968261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,32,power_law_1.2,0.037510401010513304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,32,power_law_1.01,0.17159039974212648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,32,power_law_1.2,0.03765760064125061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,32,power_law_1.01,0.20801920890808107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,32,power_law_1.2,0.037503999471664426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,32,power_law_1.01,0.3022527933120728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,32,power_law_1.2,0.03731839954853058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,32,power_law_1.2,0.03711999952793121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,32,power_law_1.01,0.6064767837524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,32,power_law_1.2,0.038387200236320494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,32,power_law_1.2,0.03866879940032959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,32,power_law_1.2,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,32,power_law_1.2,0.03916159868240356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,32,power_law_1.2,0.04281600117683411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,32,power_law_1.2,0.043110400438308716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,32,power_law_1.2,0.042931199073791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,32,power_law_1.2,0.04622719883918762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,32,power_law_1.2,0.04996480047702789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,32,power_law_1.2,0.056364798545837404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,32,power_law_1.2,0.0590399980545044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,32,power_law_1.2,0.07779840230941773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,32,power_law_1.2,0.0874239981174469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,32,power_law_1.2,0.1149440050125122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,32,power_law_1.2,0.15159039497375487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,32,power_law_1.2,0.22614400386810302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,32,power_law_1.2,0.24694399833679198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,32,power_law_1.2,0.36935040950775144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,32,power_law_1.2,0.6593215942382813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,32,power_law_1.2,1.347923183441162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,32,power_law_1.2,0.03345920145511627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,32,power_law_1.2,0.033024001121521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,32,power_law_1.2,0.032204800844192506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,32,power_law_1.2,0.03457280099391937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,32,power_law_1.2,0.03523840010166168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,32,power_law_1.2,0.03600000143051148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,32,power_law_1.2,0.0383103996515274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,32,power_law_1.2,0.03617280125617981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,32,power_law_1.2,0.037273600697517395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,32,power_law_1.2,0.037049600481987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,32,power_law_1.2,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,32,power_law_1.2,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,32,power_law_1.2,0.03903360068798065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,32,power_law_1.2,0.03950720131397247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,32,power_law_1.2,0.04267520010471344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,32,power_law_1.2,0.043647998571395875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,32,power_law_1.2,0.04552960097789764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,32,power_law_1.2,0.04997119903564453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,32,power_law_1.2,0.05631359815597534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,32,power_law_1.2,0.06083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,32,power_law_1.2,0.06782720088958741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,32,power_law_1.2,0.08321279883384705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,32,power_law_1.2,0.11266560554504394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,32,power_law_1.2,0.15544320344924928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,32,power_law_1.2,0.20972800254821777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,32,power_law_1.2,0.28983678817749026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,32,power_law_1.2,0.34286720752716066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,32,power_law_1.2,0.5430592060089111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,32,power_law_1.2,0.7716928005218506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,32,power_law_1.2,1.8251520156860352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,32,power_law_1.2,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,32,power_law_1.2,0.03214719891548157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,32,power_law_1.2,0.03258880078792572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,32,power_law_1.2,0.01674239933490753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,32,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,32,power_law_1.2,0.018278400599956512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,32,power_law_1.2,0.03640959858894348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,32,power_law_1.2,0.018009600043296815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,32,power_law_1.2,0.03927040100097656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,32,power_law_1.2,0.01865600049495697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,32,power_law_1.2,0.0389631986618042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,32,power_law_1.2,0.018214400112628936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,32,power_law_1.2,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,32,power_law_1.2,0.022035199403762817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,32,power_law_1.2,0.03948160111904144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,32,power_law_1.2,0.032876798510551454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,32,power_law_1.2,0.03925760090351105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,32,power_law_1.2,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,32,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,32,power_law_1.2,0.028966400027275085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,32,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,32,power_law_1.2,0.02910720109939575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,32,power_law_1.2,0.04084480106830597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,32,power_law_1.2,0.029407998919486998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,32,power_law_1.2,0.042745599150657655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,32,power_law_1.2,0.029574400186538695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,32,power_law_1.2,0.04565759897232056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,32,power_law_1.2,0.02895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,32,power_law_1.2,0.04689919948577881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,32,power_law_1.2,0.028780800104141236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,16,balanced,0.03182933231194814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,32,power_law_1.2,0.0486272007226944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,32,power_law_1.2,0.030611199140548707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,16,balanced,0.03299200038115183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,32,power_law_1.2,0.05188480019569397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,16,balanced,0.03418133407831192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,32,power_law_1.2,0.030694401264190672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,16,balanced,0.03679466744263967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,32,power_law_1.2,0.056524801254272464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,32,power_law_1.2,0.03364480137825012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,16,balanced,0.038431999584039055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,32,power_law_1.2,0.06323840022087097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,32,power_law_1.2,0.03320319950580597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,16,balanced,0.04137066751718521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,32,power_law_1.2,0.06908159852027893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,16,balanced,0.04193066557248434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,32,power_law_1.2,0.03658879995346069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,16,balanced,0.03847466657559077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,32,power_law_1.2,0.08714240193367004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,32,power_law_1.2,0.04094719886779785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,16,balanced,0.0408746674656868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,32,power_law_1.2,0.1018496036529541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,32,power_law_1.2,0.044556799530982974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,16,balanced,0.03846399982770284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,32,power_law_1.2,0.15117440223693848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,32,power_law_1.2,0.057657599449157715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,16,balanced,0.038506666819254555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,32,power_law_1.2,0.1871999979019165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,32,power_law_1.2,0.0657920002937317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,16,balanced,0.040463998913764954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,32,power_law_1.2,0.30960640907287595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,16,balanced,0.03841600070397059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,32,power_law_1.2,0.09125120043754578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,16,balanced,0.04037333279848099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,32,power_law_1.2,0.3660991907119751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,32,power_law_1.2,0.10971519947052003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,16,balanced,0.04436799883842468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,32,power_law_1.2,0.4465023994445801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,32,power_law_1.2,0.15838079452514647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,16,balanced,0.04206933577855428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,32,power_law_1.2,0.1919360041618347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,32,power_law_1.2,0.8399744033813477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,16,balanced,0.04418666660785675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,32,power_law_1.2,0.23210880756378174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,32,power_law_1.2,1.8337152481079102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,16,balanced,0.044362664222717285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,32,power_law_1.2,0.3432703971862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,16,balanced,0.04668800036112467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,32,power_law_1.2,0.6909632205963134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,16,balanced,0.047925333182017006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,16,balanced,0.050437331199645996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,16,balanced,0.05832533538341522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,16,balanced,0.06217066446940104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,16,balanced,0.07872533301512401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,16,balanced,0.09107733766237895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,16,balanced,0.031701333820819855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,16,balanced,0.0322026660044988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,16,balanced,0.03274133304754893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,16,balanced,0.036533333361148834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,16,balanced,0.03669866671164831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,16,balanced,0.040250666439533234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,16,balanced,0.04026666780312856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,16,balanced,0.038389332592487335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,16,balanced,0.038848000268141426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,16,balanced,0.0401706670721372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,16,balanced,0.1204800009727478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,16,balanced,0.039674667020638786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,16,balanced,0.04264533519744873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,16,balanced,0.040192000567913055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,16,balanced,0.04247466723124186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,16,balanced,0.045893331368764244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,16,balanced,0.044624000787734985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,16,balanced,0.044682666659355164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,16,balanced,0.0476746658484141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,16,balanced,0.14786666631698608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,16,balanced,0.05072533090909322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,16,balanced,0.05277333160241445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,16,balanced,0.0584853341182073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,16,balanced,0.06832533578077953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,16,balanced,0.17324266831080118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,16,balanced,0.08171199758847554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,16,balanced,0.10909866293271382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,16,balanced,0.13449600338935852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,16,balanced,0.2521546681722005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,16,balanced,0.18327999114990234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,16,balanced,0.2320586641629537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,16,power_law_1.01,0.03505280017852783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,16,power_law_1.01,0.030969598889350893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,16,power_law_1.01,0.032543998956680295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,16,power_law_1.01,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,16,power_law_1.01,0.035750401020050046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,16,power_law_1.01,0.03720319867134094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,16,balanced,0.27911466360092163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,16,power_law_1.01,0.03640959858894348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,16,power_law_1.01,0.03781760036945343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,16,power_law_1.01,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,16,power_law_1.01,0.037503999471664426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,16,power_law_1.01,0.03219200074672699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,16,power_law_1.01,0.037452799081802365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,16,power_law_1.01,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,16,power_law_1.01,0.03795199990272522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,16,power_law_1.01,0.03632639944553375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,16,power_law_1.01,0.03891839981079102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,16,power_law_1.01,0.037350401282310486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,16,power_law_1.01,0.039443200826644896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,16,power_law_1.01,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,16,power_law_1.01,0.03871999979019165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,16,power_law_1.01,0.03847039937973022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,16,power_law_1.01,0.0395583987236023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,16,power_law_1.01,0.03848319947719574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,16,power_law_1.01,0.04440320134162903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,16,power_law_1.01,0.03824000060558319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,16,power_law_1.01,0.04323840141296387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,16,balanced,0.4702346722284953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,16,power_law_1.01,0.03946239948272705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,16,power_law_1.01,0.04472320079803467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,16,power_law_1.01,0.03963519930839539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,16,power_law_1.01,0.047712001204490664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,16,power_law_1.01,0.039603200554847715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,16,power_law_1.01,0.04989440143108368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,16,power_law_1.01,0.04053759872913361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,16,power_law_1.01,0.05486720204353333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,16,power_law_1.01,0.04371199905872345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,16,power_law_1.01,0.060659199953079224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,16,power_law_1.01,0.046540799736976626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,16,power_law_1.01,0.07420160174369812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,16,power_law_1.01,0.04596480131149292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,16,power_law_1.01,0.08771839737892151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,16,power_law_1.01,0.050809597969055174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,16,power_law_1.01,0.10922880172729492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,16,balanced,0.41486934820810956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,16,power_law_1.01,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,16,power_law_1.01,0.12264319658279418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,16,power_law_1.01,0.06225280165672302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,16,power_law_1.01,0.17301759719848633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,16,power_law_1.01,0.06866559982299805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,16,power_law_1.01,0.2482624053955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,16,power_law_1.01,0.08145920038223267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,16,power_law_1.01,0.2969919919967651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,16,power_law_1.01,0.1072767972946167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,16,power_law_1.01,0.4030464172363281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,16,power_law_1.01,0.14631680250167847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,16,power_law_1.01,0.7985919952392578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,16,power_law_1.01,0.19356160163879393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,16,power_law_1.01,0.2636224031448364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,16,power_law_1.01,0.31585919857025146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,16,power_law_1.01,0.3755840063095093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,16,power_law_1.01,0.6584767818450927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,16,power_law_1.01,1.2569151878356934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,16,balanced,0.7954346338907877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,16,power_law_1.2,0.03763200044631958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,16,power_law_1.2,0.032307198643684386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,16,power_law_1.2,0.032518398761749265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,16,power_law_1.2,0.03524479866027832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,16,power_law_1.2,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,16,power_law_1.2,0.0377023994922638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,16,power_law_1.2,0.03826560080051422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,16,power_law_1.2,0.03781760036945343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,16,power_law_1.2,0.03783040046691895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,16,power_law_1.2,0.03839359879493713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,16,power_law_1.2,0.03806079924106598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,16,power_law_1.2,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,16,power_law_1.2,0.03948160111904144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,16,power_law_1.2,0.04079999923706055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,16,power_law_1.2,0.04254080057144165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,16,power_law_1.2,0.044460800290107724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,16,power_law_1.2,0.04441600143909454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,16,power_law_1.2,0.0473471999168396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,16,power_law_1.2,0.05123199820518494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,16,power_law_1.2,0.05767040252685547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,16,power_law_1.2,0.06618239879608154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,16,power_law_1.2,0.08176000118255615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,16,power_law_1.2,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,16,power_law_1.2,0.09637759923934937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,16,power_law_1.2,0.031814399361610415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,16,power_law_1.2,0.12334719896316529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,16,power_law_1.2,0.032793599367141726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,16,power_law_1.2,0.16516480445861817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,16,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,16,power_law_1.2,0.23320960998535156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,16,power_law_1.2,0.03519999980926514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,16,power_law_1.2,0.03712640106678009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,16,power_law_1.2,0.2867455959320068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,64,balanced,0.04701866706212362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,16,power_law_1.2,0.03964160084724426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,16,power_law_1.2,0.43798398971557617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,64,balanced,0.049253334601720176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,16,power_law_1.2,0.03852159976959228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,16,power_law_1.2,0.7083327770233154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,64,balanced,0.04785599807898203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,16,power_law_1.2,0.03902080059051514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,64,balanced,0.04966400067011515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,16,power_law_1.2,1.4421504020690918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,16,power_law_1.2,0.038764798641204835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,64,balanced,0.05593066910902659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,64,balanced,0.059631998340288796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,16,power_law_1.2,0.03859840035438537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,64,balanced,0.05717866619427999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,16,power_law_1.2,0.039238399267196654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,64,balanced,0.05646933118502299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,16,power_law_1.2,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,64,balanced,0.05677333474159241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,16,power_law_1.2,0.04203520119190216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,64,balanced,0.06783466537793477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,16,power_law_1.2,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,64,balanced,0.05474133292833964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,16,power_law_1.2,0.04657280147075653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,64,balanced,0.05691733459631602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,16,power_law_1.2,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,64,balanced,0.06146133442719778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,16,power_law_1.2,0.05169919729232788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,64,balanced,0.058917333682378135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,16,power_law_1.2,0.05817599892616272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,64,balanced,0.06542933483918507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,16,power_law_1.2,0.06570240259170532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,64,balanced,0.06251733501752217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,16,power_law_1.2,0.0757695972919464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,16,power_law_1.2,0.09130880236625671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,64,balanced,0.06523733337720235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,16,power_law_1.2,0.12007679939270019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,64,balanced,0.06485866506894429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,16,power_law_1.2,0.16215039491653443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,16,power_law_1.2,0.19338879585266114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,64,balanced,0.0613013356924057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,16,power_law_1.2,0.3337408065795898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,64,balanced,0.06257600088914235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,16,power_law_1.2,0.4445824146270752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,64,balanced,0.04005333284536997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,16,power_law_1.2,0.4851776123046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,64,balanced,0.04119999955097834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,64,balanced,0.06746666630109151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,64,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,16,power_law_1.2,1.0058367729187012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,64,balanced,0.043365334471066795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,64,balanced,0.05091199775536855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,16,power_law_1.2,2.2756864547729494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,64,balanced,0.046666666865348816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,64,balanced,0.07718400160471599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,64,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,64,balanced,0.04859733581542969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,64,balanced,0.049082666635513306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,64,balanced,0.04855999847253164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,64,balanced,0.04712533454100291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,64,balanced,0.083514670530955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,64,balanced,0.05156266689300537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,64,balanced,0.05068266888459524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,64,balanced,0.04867733518282572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,64,balanced,0.05271466573079427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,64,balanced,0.10389866431554158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,64,balanced,0.05333866675694784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,64,balanced,0.05284800132115682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,64,balanced,0.05425600210825602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,64,balanced,0.05667200187842051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,64,balanced,0.12379733721415202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,64,balanced,0.06266133487224579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,64,balanced,0.06878933310508728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,64,balanced,0.09320533275604248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,64,balanced,0.1529706617196401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,64,balanced,0.1097866694132487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,64,balanced,0.14653866489728293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,64,balanced,0.18992000818252563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,64,balanced,0.1793760061264038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,64,balanced,0.2290239930152893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,64,balanced,0.24002132813135782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,64,balanced,0.3044426639874776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,64,balanced,0.040175999204317726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,64,balanced,0.0424586683511734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,64,balanced,0.040565334260463715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,64,balanced,0.04372799893220266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,64,balanced,0.32738665739695233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,64,balanced,0.050767997900644936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,64,balanced,0.04894933104515076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,64,balanced,0.04878933231035868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,64,balanced,0.048453330993652344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,64,balanced,0.04884799818197886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,64,balanced,0.367850661277771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,64,balanced,0.04861866434415182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,64,balanced,0.04801600178082784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,64,balanced,0.04875733455022176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,64,balanced,0.04880533119042715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,64,balanced,0.04869333406289419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,64,balanced,0.0543146679798762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,64,balanced,0.05310399830341339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,64,balanced,0.05310933291912079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,64,balanced,0.05423999826113383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,64,balanced,0.0581226646900177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,64,balanced,0.057215998570124306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,64,balanced,0.0625439981619517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,64,balanced,0.021685334543387096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,64,balanced,0.023007998863856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,64,balanced,0.07946666578451793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,64,balanced,0.02223466585079829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,64,balanced,0.024175999065240223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,64,balanced,0.02845333268245061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,64,balanced,0.043680002291997276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,64,balanced,0.5554240147272745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,64,balanced,0.09109333157539368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,64,balanced,0.041450666884581246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,64,balanced,0.0408693328499794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,64,balanced,0.040847999354203544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,64,balanced,0.041077333192030586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,64,balanced,0.04142933338880539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,64,balanced,0.03920533259709676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,64,balanced,0.11152000228563945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,64,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,64,balanced,0.03634133438269297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,64,balanced,0.04456000030040741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,64,balanced,0.0447626660267512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,64,balanced,0.03930133332808813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,64,balanced,0.13205867012341818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,64,balanced,0.044677332043647766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,64,balanced,0.054373333851496376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,64,balanced,0.6023360093434652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,64,balanced,0.061247999469439186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,64,balanced,0.0735999991496404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,64,balanced,0.17307732502619425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,64,balanced,0.09638933340708415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,64,balanced,0.12195733189582825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,64,balanced,0.21587733427683511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,64,balanced,0.1655946671962738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,64,balanced,0.2104319930076599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,64,balanced,1.0580000082651775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,64,balanced,0.26026666164398193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,64,balanced,0.3022986650466919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,64,balanced,0.3828959862391154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,64,balanced,0.3924800157546997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,64,balanced,0.485210657119751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,64,balanced,0.7129120031992594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,64,balanced,0.7578879992167155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,64,power_law_1.01,0.050783997774124144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,64,power_law_1.01,0.053427201509475705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,64,power_law_1.01,0.04821119904518127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,64,power_law_1.01,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,64,power_law_1.01,0.055718398094177245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,64,power_law_1.01,0.05626879930496216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,64,power_law_1.01,0.054022401571273804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,64,power_law_1.01,0.05959039926528931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,64,power_law_1.01,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,64,power_law_1.01,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,64,power_law_1.01,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,64,power_law_1.01,0.053932797908782956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,64,power_law_1.01,0.05480960011482239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,64,power_law_1.01,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,64,power_law_1.01,0.05477759838104248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,64,power_law_1.01,0.058387202024459836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,64,power_law_1.01,0.06383360028266907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,64,power_law_1.01,0.07002879977226258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,64,power_law_1.01,0.07366399765014649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,64,power_law_1.01,0.09082880020141601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,64,power_law_1.01,0.10520319938659668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,64,power_law_1.01,0.13277440071105956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,64,power_law_1.01,0.155731201171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,64,power_law_1.01,0.20160000324249266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,64,power_law_1.01,0.28970239162445066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,64,power_law_1.01,0.40271358489990233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,64,power_law_1.01,0.5350848197937011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,64,power_law_1.01,0.04674560129642487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,64,power_law_1.01,0.590399980545044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,64,power_law_1.01,0.04390400052070618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,64,power_law_1.01,0.03928320109844208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,64,power_law_1.01,1.0478591918945312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,64,power_law_1.01,0.04538240134716034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,64,power_law_1.01,1.9842048645019532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,64,power_law_1.01,0.04568319916725159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,64,power_law_1.01,0.04475519955158234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,64,power_law_1.01,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,64,power_law_1.01,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,64,power_law_1.01,0.046137601137161255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,64,power_law_1.01,0.044300800561904906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,64,power_law_1.01,0.045612800121307376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,64,power_law_1.01,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,64,balanced,1.4870293935139973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,64,power_law_1.01,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,64,power_law_1.01,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,64,power_law_1.01,0.05011199712753296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,64,power_law_1.01,0.05148800015449524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,64,power_law_1.01,0.054118400812149046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,64,power_law_1.01,0.06180480122566223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,64,power_law_1.01,0.07017599940299987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,64,power_law_1.01,0.08412160277366638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,64,power_law_1.01,0.09833599925041199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,64,power_law_1.01,0.13295359611511232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,64,power_law_1.01,0.1603327989578247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,64,power_law_1.01,0.21530239582061766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,64,power_law_1.01,0.2839616060256958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,64,power_law_1.01,0.40891518592834475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,64,power_law_1.01,0.5216896057128906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,64,power_law_1.01,0.6418047904968261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,64,power_law_1.01,1.0568127632141113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,64,power_law_1.01,2.0238975524902343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,64,power_law_1.01,0.04309119880199432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,64,power_law_1.01,0.04239999949932098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,64,power_law_1.01,0.04139519929885864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,64,power_law_1.01,0.046387198567390445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,64,power_law_1.01,0.04689919948577881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,64,power_law_1.01,0.04749439954757691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,64,power_law_1.01,0.046060800552368164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,64,power_law_1.01,0.04584319889545441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,64,power_law_1.01,0.04673280119895935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,64,power_law_1.01,0.04672000110149384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,64,power_law_1.01,0.047417598962783816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,64,power_law_1.01,0.04749439954757691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,64,power_law_1.01,0.04739840030670166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,64,power_law_1.01,0.047417598962783816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,64,power_law_1.01,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,64,power_law_1.01,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,64,power_law_1.01,0.055852800607681274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,64,power_law_1.01,0.05861120223999024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.01,0.06848639845848084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.01,0.07594239711761475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.01,0.08910080194473266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.01,0.10892159938812256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.01,0.12932480573654176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.01,0.1918463945388794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.01,0.22190721035003663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.01,0.31217920780181885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.01,0.4070079803466797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,64,power_law_1.01,0.48383359909057616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,64,power_law_1.01,0.7446271896362304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,64,power_law_1.01,1.4464768409729003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,64,power_law_1.01,0.020857599377632142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,64,power_law_1.01,0.020908799767494202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,64,power_law_1.01,0.01991039961576462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,64,power_law_1.01,0.021414400637149812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,64,power_law_1.01,0.02563839852809906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,64,power_law_1.01,0.040659201145172116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,64,power_law_1.2,0.04837760031223297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,64,power_law_1.01,0.03937920033931732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,64,power_law_1.2,0.048537600040435794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,64,power_law_1.01,0.04015359878540039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,64,power_law_1.2,0.04960640072822571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,64,power_law_1.01,0.040057599544525146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,64,power_law_1.2,0.05136640071868896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,64,power_law_1.01,0.03979519903659821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,64,power_law_1.2,0.052134400606155394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,64,power_law_1.01,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,64,power_law_1.2,0.055430400371551516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,64,power_law_1.01,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,64,power_law_1.2,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,64,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,64,power_law_1.01,0.035596799850463864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,64,power_law_1.2,0.05470079779624939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,64,power_law_1.01,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,64,power_law_1.2,0.05451520085334778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,64,power_law_1.2,0.05248000025749207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,64,power_law_1.01,0.037215998768806456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,64,power_law_1.2,0.05432959794998169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,64,power_law_1.01,0.03949440121650696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,64,power_law_1.2,0.05167359709739685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,64,power_law_1.01,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,64,power_law_1.2,0.05151360034942627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.01,0.04907520115375519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,64,power_law_1.2,0.05434880256652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.01,0.05925760269165039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,64,power_law_1.2,0.05750399827957153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.01,0.06547840237617493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,64,power_law_1.2,0.06380800008773804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.01,0.08692479729652405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,64,power_law_1.2,0.06385279893875122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.01,0.10572160482406616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,64,power_law_1.2,0.08067839741706848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.01,0.13964799642562867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,64,power_law_1.2,0.08507519960403442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.01,0.17177599668502808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,64,power_law_1.2,0.1121216058731079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.01,0.27407360076904297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,64,power_law_1.2,0.11960320472717285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.01,0.3558464050292969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,64,power_law_1.2,0.16933759450912475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,64,power_law_1.01,0.43283839225769044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,64,power_law_1.2,0.047219198942184445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,64,power_law_1.2,0.19738240242004396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,64,power_law_1.01,0.7076352119445801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,64,power_law_1.2,0.04160000085830688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,64,power_law_1.2,0.23796479701995848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,64,power_law_1.2,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,64,power_law_1.2,0.41446399688720703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,64,power_law_1.01,1.3467967987060547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,64,power_law_1.2,0.043808001279830935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,64,power_law_1.2,0.6412415981292725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,64,power_law_1.2,0.04347519874572754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,64,power_law_1.2,0.6997375965118409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,64,power_law_1.2,0.04530560076236725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,64,power_law_1.2,1.1538304328918456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,64,power_law_1.2,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,64,power_law_1.2,1.9863935470581056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,64,power_law_1.2,0.04748159945011139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,64,power_law_1.2,0.04565759897232056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,64,power_law_1.2,3.1342016220092774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,64,power_law_1.2,0.04517120122909546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,64,power_law_1.2,0.04656000137329101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,64,power_law_1.2,0.04673280119895935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,64,power_law_1.2,0.048147198557853696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,64,power_law_1.2,0.04668160080909729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,64,power_law_1.2,0.05383679866790771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,64,power_law_1.2,0.053446400165557864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,64,power_law_1.2,0.05603839755058289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,64,power_law_1.2,0.06472319960594178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,64,power_law_1.2,0.07728000283241272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,64,power_law_1.2,0.09587200284004212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,64,power_law_1.2,0.11203199625015259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,64,power_law_1.2,0.15002880096435547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,64,power_law_1.2,0.1790719985961914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,64,power_law_1.2,0.2569792032241821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,64,power_law_1.2,0.42780160903930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,64,power_law_1.2,0.6729343891143799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,64,power_law_1.2,0.6933119773864747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,64,power_law_1.2,0.8831616401672363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,64,power_law_1.2,1.6930047988891601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,64,power_law_1.2,3.0768320083618166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,64,power_law_1.2,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,64,power_law_1.2,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,64,power_law_1.2,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,64,power_law_1.2,0.0441536009311676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,64,power_law_1.2,0.04509440064430237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,64,power_law_1.2,0.046374401450157164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,64,power_law_1.2,0.045900800824165346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,64,power_law_1.2,0.046035200357437134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,64,power_law_1.2,0.04642559885978699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,64,power_law_1.2,0.04717440009117126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,64,power_law_1.2,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,64,power_law_1.2,0.04630399942398071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,64,power_law_1.2,0.04633600115776062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,64,power_law_1.2,0.04746879935264588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,64,power_law_1.2,0.052153599262237546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,64,power_law_1.2,0.05504639744758606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,64,power_law_1.2,0.060108798742294314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,64,power_law_1.2,0.06386560201644897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.2,0.07057920098304749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.2,0.08485119938850402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.2,0.0998080015182495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.2,0.1293951988220215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.2,0.16152960062026978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.2,0.21782400608062744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.2,0.28897919654846194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.2,0.4024511814117432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.2,0.6632063865661622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,64,power_law_1.2,0.8321855545043946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,64,power_law_1.2,1.5094271659851075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,64,power_law_1.2,2.655417633056641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,64,power_law_1.2,0.020921599864959717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,64,power_law_1.2,0.02110079973936081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,64,power_law_1.2,0.02024320065975189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,32,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,64,power_law_1.2,0.020051200687885285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,32,balanced,0.040933333337306976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,64,power_law_1.2,0.026348799467086792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,32,balanced,0.04079466561476389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,64,power_law_1.2,0.04046080112457275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,32,balanced,0.046282668908437095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,64,power_law_1.2,0.03903360068798065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,32,balanced,0.054799998799959816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,32,balanced,0.051088000337282814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,64,power_law_1.2,0.04043520092964172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,32,balanced,0.05727999905745188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,64,power_law_1.2,0.042131200432777405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,32,balanced,0.053183997670809426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,64,power_law_1.2,0.03925760090351105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,32,balanced,0.05749333401521047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,64,power_law_1.2,0.040012800693511964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,32,balanced,0.06117333471775055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,64,power_law_1.2,0.03891839981079102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,32,balanced,0.05528533458709717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,32,balanced,0.057114665706952415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,64,power_law_1.2,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,32,balanced,0.05890133480230967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,64,power_law_1.2,0.03407999873161316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,32,balanced,0.058634668588638306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,64,power_law_1.2,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,32,balanced,0.05859733124574026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,64,power_law_1.2,0.0369024008512497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,32,balanced,0.059215997656186424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,64,power_law_1.2,0.03943040072917938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,64,power_law_1.2,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,32,balanced,0.05884799857934316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.2,0.051520001888275144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,32,balanced,0.05756799876689911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.2,0.05939840078353882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.2,0.07127040028572082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,32,balanced,0.057445332407951355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.2,0.09201920032501221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,32,balanced,0.06277333199977875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.2,0.11292799711227416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.2,0.160588800907135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.2,0.19879039525985717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,32,balanced,0.06706133484840393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.2,0.28769280910491946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.2,0.38407680988311765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,32,balanced,0.08123200138409932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,64,power_law_1.2,0.4609536170959473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,64,power_law_1.2,0.7335423946380615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,64,power_law_1.2,1.4509119987487793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,32,balanced,0.08959999680519104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,32,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,32,balanced,0.03676266719897588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,32,balanced,0.036464000741640724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,32,balanced,0.04311466713746389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,32,balanced,0.04533333579699198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,32,balanced,0.10770666599273682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,32,balanced,0.049509331583976746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,32,balanced,0.04446933170159658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,32,balanced,0.04900266726811727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,32,balanced,0.048842668533325195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,32,balanced,0.04915733138720194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,32,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,32,balanced,0.04880533119042715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,32,balanced,0.048997332652409874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,32,balanced,0.13499733805656433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,32,balanced,0.050704002380371094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,32,balanced,0.05053333441416422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,32,balanced,0.05118933320045471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,32,balanced,0.05494399865468343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,32,balanced,0.055402666330337524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,32,balanced,0.05855466425418854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,32,balanced,0.06365333497524261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,32,balanced,0.16370667020479837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,32,balanced,0.07273066540559132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,32,balanced,0.10003200173377991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,32,balanced,0.11562666296958923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,32,balanced,0.2042186657587687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,32,balanced,0.1529973347981771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,32,balanced,0.0377813329299291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,32,balanced,0.03877866764863332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,32,balanced,0.18859734137852988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,32,balanced,0.03871466716130575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,32,balanced,0.04309333364168803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,32,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,32,balanced,0.05086933573087057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,32,balanced,0.051221330960591636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,32,balanced,0.048309331138928734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,32,balanced,0.05082666873931885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,32,balanced,0.050053333242734276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,32,balanced,0.24624532461166382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,32,balanced,0.05050133168697357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,32,balanced,0.04911999901135763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,32,balanced,0.2550400098164876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,32,balanced,0.04882133503754934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,32,balanced,0.04889066517353058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,32,balanced,0.05538133283456167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,32,balanced,0.056736002365748085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,32,balanced,0.05481599768002828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,32,balanced,0.05671999851862589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,32,balanced,0.06080000102519989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,32,balanced,0.06228800117969513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,32,balanced,0.3238933285077413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,32,balanced,0.06758399804433186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,32,balanced,0.08410666386286418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,32,balanced,0.09762666622797649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,32,balanced,0.3590293327967326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,32,balanced,0.12027733524640401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,32,balanced,0.39163732528686523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,32,balanced,0.1414400041103363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,32,balanced,0.18894400199254355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,32,balanced,0.021125334004561108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,32,balanced,0.02195200075705846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,32,balanced,0.02184533327817917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,32,balanced,0.021877333521842957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,32,balanced,0.5978879928588867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,32,balanced,0.023584000766277313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,32,balanced,0.02664533257484436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,32,balanced,0.04301333427429199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,32,balanced,0.23691733678181967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,32,balanced,0.043354665239652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,32,balanced,0.04081066697835922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,32,balanced,0.0405973345041275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,32,balanced,0.040405333042144775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,32,balanced,0.03914133210976919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,32,balanced,0.03682666768630346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,32,balanced,0.0359253336985906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,32,balanced,0.0470719983180364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,32,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,32,balanced,0.042266666889190674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,32,balanced,0.044693330923716225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,32,balanced,0.056874667604764305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,32,balanced,0.28223999341328937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,32,balanced,0.06699733436107635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,32,balanced,0.6641173362731934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,32,balanced,0.08066133161385854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,32,balanced,0.10409599542617798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,32,balanced,0.13038399815559387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,32,balanced,0.17511467138926187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,32,balanced,0.4195733467737834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,32,balanced,0.2246933380762736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,32,balanced,1.1467466354370117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,32,balanced,0.3202986717224121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,32,balanced,0.4134720166524251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,32,balanced,0.5125866731007894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,32,balanced,0.7870559692382812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,32,balanced,0.8157280286153158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,32,power_law_1.01,0.04899199903011322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,32,power_law_1.01,0.04410879909992218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,32,power_law_1.01,0.04071680009365082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,32,power_law_1.01,0.048179200291633605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,32,power_law_1.01,0.04915199875831604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,32,power_law_1.01,0.051596802473068235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,32,power_law_1.01,0.05160319805145264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,32,power_law_1.01,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,32,power_law_1.01,0.051526397466659546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,32,power_law_1.01,0.05082880258560181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,32,power_law_1.01,0.05231999754905701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,32,power_law_1.01,0.05278720259666443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,32,power_law_1.01,0.05135999917984009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,32,power_law_1.01,0.051532799005508424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,32,power_law_1.01,0.05480960011482239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,32,power_law_1.01,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,32,power_law_1.01,0.06529920101165772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,32,power_law_1.01,0.06609920263290406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,32,power_law_1.01,0.07294719815254211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,32,power_law_1.01,0.08240000009536744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,32,power_law_1.01,0.09676160216331482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,32,power_law_1.01,0.12620799541473388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,32,power_law_1.01,0.13534719944000245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,32,balanced,1.6010133425394695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,32,power_law_1.01,0.1844607949256897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,32,power_law_1.01,0.22744319438934327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,32,power_law_1.01,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,32,power_law_1.01,0.3026432037353516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,32,power_law_1.01,0.03654400110244751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,32,power_law_1.01,0.4410111904144287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,32,power_law_1.01,0.0367935985326767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,32,power_law_1.01,0.5410496234893799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,32,power_law_1.01,0.041254401206970215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,32,power_law_1.01,0.8039999961853027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,32,power_law_1.01,0.042054399847984314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,32,power_law_1.01,0.04583680033683777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,32,power_law_1.01,1.481004810333252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,32,power_law_1.01,0.044889599084854126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,32,power_law_1.01,0.04792959988117218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,32,power_law_1.01,0.0443583995103836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,32,power_law_1.01,0.045772799849510194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,32,power_law_1.01,0.04672000110149384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,32,power_law_1.01,0.046982398629188536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,32,power_law_1.01,0.047040000557899475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,32,power_law_1.01,0.04687359929084778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,32,power_law_1.01,0.0510591983795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,32,power_law_1.01,0.05112959742546082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,32,power_law_1.01,0.054841601848602296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,32,power_law_1.01,0.058796799182891844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,32,power_law_1.01,0.0671999990940094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,32,power_law_1.01,0.0813759982585907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,32,power_law_1.01,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,32,power_law_1.01,0.10147199630737305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,32,power_law_1.01,0.04390400052070618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,32,power_law_1.01,0.12696319818496704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,32,power_law_1.01,0.039052799344062805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,32,power_law_1.01,0.15946880578994752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,32,power_law_1.01,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,32,power_law_1.01,0.044300800561904906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,32,power_law_1.01,0.20885119438171387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,32,power_law_1.01,0.04612480103969574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,32,power_law_1.01,0.25478401184082033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,32,power_law_1.01,0.047865599393844604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,32,power_law_1.01,0.3650815963745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,32,power_law_1.01,0.04752640128135681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,32,power_law_1.01,0.5146880149841309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,32,power_law_1.01,0.04663679897785187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,32,power_law_1.01,0.6087423801422119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,32,power_law_1.01,0.047513601183891294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,32,power_law_1.01,1.0347519874572755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,32,power_law_1.01,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,32,power_law_1.01,1.9279167175292968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,32,power_law_1.01,0.04863359928131104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,32,power_law_1.01,0.049235200881958006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,32,power_law_1.01,0.04943360090255737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,32,power_law_1.01,0.053497600555419925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,32,power_law_1.01,0.054092800617218016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,32,power_law_1.01,0.05616000294685364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,32,power_law_1.01,0.06079999804496765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.01,0.06871680021286011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.01,0.07803519964218139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.01,0.08831359744071961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.01,0.11128319501876831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.01,0.12577919960021972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.01,0.18469120264053346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.01,0.2065984010696411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.01,0.31154561042785645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.01,0.40844159126281737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,32,power_law_1.01,0.49923200607299806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,32,power_law_1.01,0.8661375999450683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,32,power_law_1.01,1.4852992057800294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,32,power_law_1.01,0.018911999464035035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,32,power_law_1.01,0.019699199497699736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,32,power_law_1.01,0.019808000326156615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,32,power_law_1.01,0.019577600061893463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,32,power_law_1.01,0.02170239984989166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,32,power_law_1.01,0.026604801416397095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,32,power_law_1.2,0.05248000025749207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,32,power_law_1.01,0.04083200097084046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,32,power_law_1.2,0.04090240001678467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,32,power_law_1.01,0.041068801283836366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,32,power_law_1.2,0.04570879936218262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,32,power_law_1.01,0.03949440121650696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,32,power_law_1.2,0.04560000002384186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,32,power_law_1.01,0.0392192006111145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,32,power_law_1.2,0.04799999892711639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,32,power_law_1.01,0.03937279880046844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,32,power_law_1.2,0.05305600166320801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,32,power_law_1.2,0.05260800123214722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,32,power_law_1.01,0.03999359905719757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,32,power_law_1.2,0.05426560044288635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,32,power_law_1.01,0.038022398948669434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,32,power_law_1.2,0.04990079998970032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,32,power_law_1.01,0.0382207989692688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,32,power_law_1.2,0.052748799324035645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,32,power_law_1.01,0.03976320028305054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,32,power_law_1.2,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,32,power_law_1.01,0.038764798641204835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,32,power_law_1.2,0.05219200253486633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,32,power_law_1.01,0.04418559968471527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,32,power_law_1.2,0.05390080213546753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,32,power_law_1.01,0.041388800740242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,32,power_law_1.2,0.052857601642608644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.01,0.044223999977111815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,32,power_law_1.2,0.061504000425338747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.01,0.05616639852523804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,32,power_law_1.2,0.06247680187225342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.01,0.06332160234451294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,32,power_law_1.2,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.01,0.0815168023109436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,32,power_law_1.2,0.06888960003852844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.01,0.09859200119972229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,32,power_law_1.2,0.077920001745224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.01,0.1266495943069458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,32,power_law_1.2,0.09392639994621277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.01,0.1681023955345154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,32,power_law_1.2,0.10122239589691162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.01,0.23493759632110595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,32,power_law_1.2,0.15007359981536866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.01,0.3020031929016113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,32,power_law_1.2,0.1543295979499817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,32,power_law_1.01,0.35425920486450196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,32,power_law_1.2,0.26757121086120605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,32,power_law_1.01,0.5514495849609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,32,power_law_1.2,0.28107519149780275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,32,power_law_1.01,1.2107647895812987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,32,power_law_1.2,0.4081535816192627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,32,power_law_1.2,0.0412992000579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,32,power_law_1.2,0.5732160091400147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,32,power_law_1.2,0.03686400055885315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,32,power_law_1.2,0.8438655853271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,32,power_law_1.2,0.03943040072917938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,32,power_law_1.2,0.04012799859046936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,32,power_law_1.2,1.4119296073913574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,32,power_law_1.2,0.04232319891452789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,32,power_law_1.2,4.248953628540039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,32,power_law_1.2,0.045484799146652224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,32,power_law_1.2,0.04474239945411682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,32,power_law_1.2,0.045561599731445315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,32,power_law_1.2,0.04534400105476379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,32,power_law_1.2,0.045433598756790164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,32,power_law_1.2,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,32,power_law_1.2,0.04616959989070892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,32,power_law_1.2,0.04849919974803925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,32,power_law_1.2,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,32,power_law_1.2,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,32,power_law_1.2,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,32,power_law_1.2,0.04496639966964722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,32,power_law_1.2,0.057183998823165896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,32,power_law_1.2,0.04042240083217621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,32,power_law_1.2,0.06354560256004334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,32,power_law_1.2,0.04161919951438904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,32,power_law_1.2,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,32,power_law_1.2,0.0426367998123169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,32,power_law_1.2,0.09288319945335388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,32,power_law_1.2,0.04359039962291718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,32,power_law_1.2,0.11934720277786255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,32,power_law_1.2,0.047219198942184445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,32,power_law_1.2,0.13907840251922607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,32,power_law_1.2,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,32,power_law_1.2,0.1668544054031372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,32,power_law_1.2,0.04693120121955872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,32,power_law_1.2,0.23719680309295654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,32,power_law_1.2,0.04731520116329193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,32,power_law_1.2,0.3376512050628662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,32,power_law_1.2,0.04746879935264588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,32,power_law_1.2,0.5670591831207276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,32,power_law_1.2,0.04779520034790039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,32,power_law_1.2,0.7071296215057373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,32,power_law_1.2,0.04986239969730377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,32,power_law_1.2,0.8860287666320801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,32,power_law_1.2,0.04997760057449341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,32,power_law_1.2,1.2673855781555177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,32,power_law_1.2,0.0502016007900238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,32,power_law_1.2,0.05451520085334778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,32,power_law_1.2,3.238111877441406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,32,power_law_1.2,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,32,power_law_1.2,0.06026880145072937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,32,power_law_1.2,0.06855679750442505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.2,0.07175040245056152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.2,0.09157760143280029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.2,0.09586560130119323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.2,0.12290560007095337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.2,0.15691519975662233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.2,0.21527678966522218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.2,0.2996927976608276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.2,0.40320639610290526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.2,0.5282879829406738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,32,power_law_1.2,0.7423168182373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,32,power_law_1.2,1.2380096435546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,32,power_law_1.2,2.147667121887207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,16,balanced,0.03809066613515218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,16,balanced,0.03867200016975403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,16,balanced,0.04091733445723852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,32,power_law_1.2,0.01937279999256134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,16,balanced,0.045279999574025474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,16,balanced,0.05131733417510986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,32,power_law_1.2,0.020083199441432952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,16,balanced,0.05529066423575083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,32,power_law_1.2,0.01958400011062622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,16,balanced,0.05489600201447805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,32,power_law_1.2,0.02025599926710129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,16,balanced,0.049546668926874794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,32,power_law_1.2,0.02190079987049103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,16,balanced,0.055434669057528176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,32,power_law_1.2,0.02677760124206543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,16,balanced,0.05329066514968872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,32,power_law_1.2,0.040780800580978396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,16,balanced,0.05113600194454193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,32,power_law_1.2,0.041440001130104064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,16,balanced,0.05465066432952881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,32,power_law_1.2,0.040243199467658995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,16,balanced,0.050842667619387306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,32,power_law_1.2,0.04007039964199066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,16,balanced,0.05287999908129374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,32,power_law_1.2,0.039443200826644896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,16,balanced,0.05682666599750519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,32,power_law_1.2,0.03909760117530823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,16,balanced,0.056330665946006775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,32,power_law_1.2,0.03933440148830414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,16,balanced,0.05904533465703329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,32,power_law_1.2,0.03708159923553467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,16,balanced,0.059194669127464294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,32,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,32,power_law_1.2,0.03696640133857727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,16,balanced,0.06398400167624156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,32,power_law_1.2,0.041740798950195314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,32,power_law_1.2,0.04216960072517395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,16,balanced,0.06878933310508728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.2,0.04529919922351837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.2,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,16,balanced,0.07520533104737599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.2,0.06821759939193725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.2,0.08791679739952088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.2,0.10382720232009887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,16,balanced,0.09061333537101746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.2,0.14519040584564208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.2,0.17704319953918457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.2,0.2667072057723999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,16,balanced,0.09918933113416036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.2,0.33413119316101075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,32,power_law_1.2,0.4029439926147461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,32,power_law_1.2,0.70830078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,16,balanced,0.03524799893299738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,16,balanced,0.03601066768169403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,32,power_law_1.2,1.432812786102295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,16,balanced,0.0383840004603068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,16,balanced,0.041082667807737984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,16,balanced,0.12199466427167256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,16,balanced,0.04500266909599304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,16,balanced,0.04654933512210846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,16,balanced,0.047135998805363975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,16,balanced,0.046666666865348816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,16,balanced,0.048901334404945374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,16,balanced,0.048623998959859215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,16,balanced,0.04615999758243561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,16,balanced,0.14706666270891824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,16,balanced,0.04916266600290934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,16,balanced,0.05052266518274943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,16,balanced,0.03600533306598663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,16,balanced,0.03905600061019262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,16,balanced,0.04925866425037384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,16,balanced,0.0391839991013209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,16,balanced,0.05106666684150696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,16,balanced,0.04642133414745331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,16,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,16,balanced,0.052853330969810486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,16,balanced,0.050853331883748375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,16,balanced,0.05436799923578898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,16,balanced,0.05221866567929586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,16,balanced,0.049173335234324135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,16,balanced,0.05750933289527893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,16,balanced,0.18973867098490396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,16,balanced,0.05012266834576925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,16,balanced,0.05077866713205973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,16,balanced,0.05862933397293091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,16,balanced,0.052709331115086876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,16,balanced,0.05439466734727224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,16,balanced,0.0690826674302419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,16,balanced,0.05083199838797251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,16,balanced,0.050800000627835594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,16,balanced,0.08124266564846039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,16,balanced,0.05677333474159241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,16,balanced,0.054917335510253906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,16,balanced,0.056554665168126426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,16,balanced,0.10958400368690491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,16,balanced,0.24111467599868774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,16,balanced,0.05904533465703329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,16,balanced,0.06316266457239787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,16,balanced,0.12628266215324402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,16,balanced,0.0666293352842331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,16,balanced,0.0761706680059433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,16,balanced,0.16884799798329672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,16,balanced,0.09777599573135376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,16,balanced,0.290282666683197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,16,balanced,0.11326400438944499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,16,balanced,0.21025600035985312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,16,balanced,0.14075199762980142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,16,balanced,0.2849173347155253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,16,balanced,0.1698346734046936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,16,balanced,0.43087466557820636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,16,balanced,0.3641546567281087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,16,balanced,0.22914133469263712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,16,balanced,0.2881653308868408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,16,balanced,0.44364798069000244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,16,balanced,0.3523840109507243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,16,balanced,0.8066559632619222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,16,balanced,0.6844106515248617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,16,balanced,0.02126399924357732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,16,balanced,0.022287999590237934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,16,balanced,0.020725333442290623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,16,balanced,0.02179733415444692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,16,balanced,0.022437334060668945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,16,balanced,0.5280266602834066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,16,balanced,0.023919999599456787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,16,balanced,0.02828266719977061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,16,balanced,0.02624533325433731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,16,balanced,0.04251733422279358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,16,balanced,0.042917331059773765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,16,balanced,0.044437333941459656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,16,balanced,0.043285335103670754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,16,balanced,0.04229333500067393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,16,balanced,0.04038399954636892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,16,balanced,0.04758933186531067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,16,balanced,0.04741866886615753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,16,balanced,0.04651199777921041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,16,balanced,0.05008533100287119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,16,balanced,0.05644266804059347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,16,balanced,0.07134399811426799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,16,balanced,0.07832000156243642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,16,balanced,0.09945600231488545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,16,balanced,0.11408533652623494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,16,balanced,1.3220853010813396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,16,balanced,0.16245333353678384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,16,balanced,1.0114400386810303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,16,balanced,0.19790399074554443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,16,balanced,0.2825546662012736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,16,balanced,0.3652000029881795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,16,balanced,0.4486453135808309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,16,power_law_1.01,0.0510591983795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,16,power_law_1.01,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,16,power_law_1.01,0.04051840007305145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,16,power_law_1.01,0.04565759897232056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,16,power_law_1.01,0.047167998552322385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,16,power_law_1.01,0.052902400493621826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,16,power_law_1.01,0.05184000134468079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,16,power_law_1.01,0.05203840136528015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,16,power_law_1.01,0.05123839974403381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,16,power_law_1.01,0.05130879878997803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,16,power_law_1.01,0.05242879986763001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,16,power_law_1.01,0.0525056004524231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,16,power_law_1.01,0.05375999808311462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,16,power_law_1.01,0.05489280223846436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,16,power_law_1.01,0.05848960280418396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,16,power_law_1.01,0.060134398937225345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,16,power_law_1.01,0.06166399717330932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,16,power_law_1.01,0.06661760210990905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,16,power_law_1.01,0.07328640222549439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,16,power_law_1.01,0.08229119777679443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,16,balanced,0.711413304011027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,16,power_law_1.01,0.0938368022441864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,16,power_law_1.01,0.11773439645767211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,16,power_law_1.01,0.1431167960166931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,16,power_law_1.01,0.190009605884552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,16,power_law_1.01,0.21466240882873536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,16,power_law_1.01,0.31804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,16,power_law_1.01,0.4068863868713379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,16,power_law_1.01,0.5127039909362793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,16,power_law_1.01,0.8447744369506835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,16,power_law_1.01,1.4467071533203124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,16,power_law_1.01,0.046828800439834596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,16,power_law_1.01,0.039110401272773744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,16,power_law_1.01,0.03842560052871704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,16,power_law_1.01,0.0401856005191803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,16,power_law_1.01,0.05315840244293213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,16,power_law_1.01,0.042368000745773314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,16,power_law_1.01,0.039443200826644896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,16,power_law_1.01,0.04470399916172028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,16,power_law_1.01,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,16,power_law_1.01,0.045440000295639035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,16,power_law_1.01,0.04524160027503967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,16,power_law_1.01,0.04673919975757599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,16,power_law_1.01,0.045433598756790164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,16,power_law_1.01,0.04689280092716217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,16,power_law_1.01,0.05077760219573975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,16,power_law_1.01,0.04723199903964996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,16,power_law_1.01,0.05102720260620117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,16,power_law_1.01,0.047065600752830505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,16,power_law_1.01,0.04832000136375427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,16,power_law_1.01,0.04880000054836273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,16,power_law_1.01,0.049670401215553286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,16,power_law_1.01,0.04880639910697937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,16,power_law_1.01,0.04991360008716583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,16,power_law_1.01,0.04919039905071258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,16,power_law_1.01,0.05018879771232605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,16,power_law_1.01,0.053439998626708986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,16,power_law_1.01,0.051641601324081424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,16,power_law_1.01,0.05533440113067627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,16,power_law_1.01,0.05175679922103882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,16,power_law_1.01,0.05686399936676025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,16,power_law_1.01,0.052262401580810545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,16,power_law_1.01,0.06299520134925843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,16,power_law_1.01,0.057529598474502563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,16,power_law_1.01,0.07082239985466003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,16,power_law_1.01,0.08664320111274719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,16,power_law_1.01,0.10598399639129638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,16,power_law_1.01,0.13460479974746703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,16,power_law_1.01,0.15719679594039918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,16,power_law_1.01,0.219596791267395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,16,balanced,1.3948639233907063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,16,power_law_1.01,0.27674880027771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,16,power_law_1.01,0.40133118629455566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,16,power_law_1.01,0.5594944000244141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,16,power_law_1.01,0.6356736183166504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,16,power_law_1.01,1.084832000732422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,16,power_law_1.01,2.0581056594848635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,16,power_law_1.01,0.01875839978456497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,16,power_law_1.01,0.018464000523090364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,16,power_law_1.01,0.01849599927663803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,16,power_law_1.01,0.01990399956703186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,16,power_law_1.01,0.01961600035429001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,16,power_law_1.01,0.022201600670814513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,16,power_law_1.01,0.02625280022621155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,16,power_law_1.01,0.025939199328422546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,16,power_law_1.01,0.04224640130996704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,16,power_law_1.01,0.04215039908885956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,16,power_law_1.01,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,16,power_law_1.01,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,16,power_law_1.01,0.04170880019664765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,16,power_law_1.01,0.0400191992521286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,16,power_law_1.01,0.040608000755310056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,16,power_law_1.01,0.04255360066890716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,16,power_law_1.01,0.04353919923305512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,16,power_law_1.01,0.04509440064430237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.01,0.04624640047550201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.01,0.053523200750350955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.01,0.060063999891281125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.01,0.08127359747886657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.01,0.09255679845809936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.01,0.12348159551620483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.01,0.16119680404663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.01,0.21571199893951415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.01,0.27688961029052733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,16,power_law_1.01,0.3414207935333252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,16,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,16,power_law_1.01,0.5172671794891357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,16,power_law_1.2,0.04042240083217621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,16,power_law_1.2,0.042105600237846375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,16,power_law_1.01,1.0230463981628417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,16,power_law_1.2,0.04679040014743805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,16,power_law_1.2,0.04860160052776337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,16,power_law_1.2,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,16,power_law_1.2,0.051744002103805545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,16,power_law_1.2,0.052716797590255736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,16,power_law_1.2,0.05248000025749207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,16,power_law_1.2,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,16,power_law_1.2,0.047276800870895384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,16,power_law_1.2,0.054176002740859985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,16,power_law_1.2,0.038675200939178464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,16,power_law_1.2,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,16,power_law_1.2,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,16,power_law_1.2,0.0531328022480011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,16,power_law_1.2,0.04084480106830597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,16,power_law_1.2,0.05575039982795715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,16,power_law_1.2,0.04192639887332916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,16,power_law_1.2,0.04656000137329101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,16,power_law_1.2,0.061408001184463504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,16,power_law_1.2,0.04596480131149292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,16,power_law_1.2,0.05862399935722351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,16,power_law_1.2,0.04531840085983276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,16,power_law_1.2,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,16,power_law_1.01,0.057004797458648684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,16,power_law_1.2,0.04739840030670166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,16,power_law_1.01,0.06179839968681335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,16,power_law_1.2,0.047391998767852786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,16,power_law_1.01,0.06653439998626709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,16,power_law_1.2,0.04906240105628967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.01,0.07454079985618592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,16,power_law_1.2,0.04984959959983826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.01,0.08578559756278992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,16,power_law_1.2,0.04878720045089722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.01,0.09708160161972046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,16,power_law_1.2,0.05479679703712463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,16,power_law_1.2,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.01,0.12755839824676513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,16,power_law_1.2,0.05731840133666992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.01,0.1525056004524231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,16,power_law_1.2,0.0662015974521637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.01,0.20087039470672607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,16,power_law_1.2,0.07466239929199218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.01,0.25680639743804934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,16,power_law_1.2,0.09609599709510804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.01,0.34357759952545164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,16,power_law_1.2,0.11585919857025147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.01,0.44583678245544434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,16,power_law_1.2,0.15431679487228395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,16,power_law_1.01,0.5796095848083496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,16,power_law_1.2,0.1931712031364441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,16,power_law_1.01,0.8761024475097656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,16,power_law_1.2,0.24626560211181642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,16,power_law_1.2,0.31067519187927245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,16,power_law_1.01,1.8085248947143555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,16,power_law_1.2,0.4533952236175537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,16,power_law_1.2,0.5904831886291504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,16,power_law_1.2,0.8379199981689454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,16,power_law_1.2,1.2010944366455079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,16,power_law_1.2,2.317683219909668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,16,power_law_1.2,0.04973439872264862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,16,power_law_1.2,0.037503999471664426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,16,power_law_1.2,0.0395904004573822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,16,power_law_1.2,0.043968001008033754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,16,power_law_1.2,0.046323201060295104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,16,power_law_1.2,0.04952960014343262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,16,power_law_1.2,0.049779200553894044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,16,power_law_1.2,0.050783997774124144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,16,power_law_1.2,0.04958719909191132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,16,power_law_1.2,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,16,power_law_1.2,0.05018240213394165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,16,power_law_1.2,0.05095040202140808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,16,power_law_1.2,0.05146239995956421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,16,power_law_1.2,0.05219200253486633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,16,power_law_1.2,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,16,power_law_1.2,0.05888640284538269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,16,power_law_1.2,0.06172159910202026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,16,power_law_1.2,0.06598399877548218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.2,0.07672960162162781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.2,0.09466879963874816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.2,0.10549759864807129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.2,0.13295359611511232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.2,0.16576000452041625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.2,0.2195647954940796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.2,0.33062400817871096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.2,0.47084798812866213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.2,0.6788608074188233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,16,power_law_1.2,0.6763904094696045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,16,power_law_1.2,1.112992000579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,16,power_law_1.2,2.8066944122314452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,16,power_law_1.2,0.018796800076961516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,16,power_law_1.2,0.06516479849815368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,16,power_law_1.2,0.019232000410556793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,16,power_law_1.2,0.0713919997215271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,16,power_law_1.2,0.019679999351501463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,16,power_law_1.2,0.07854080200195312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,16,power_law_1.2,0.020160000026226043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,16,power_law_1.2,0.0911296010017395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,8,balanced,0.03811733424663544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,16,power_law_1.2,0.020153599977493285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,8,balanced,0.0390133336186409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,16,power_law_1.2,0.10992640256881714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,8,balanced,0.04140799989302953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,16,power_law_1.2,0.02190079987049103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,16,power_law_1.2,0.14283519983291626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,8,balanced,0.04537599782148997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,16,power_law_1.2,0.02595840096473694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,16,power_law_1.2,0.14545919895172119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,8,balanced,0.05522133409976959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,16,power_law_1.2,0.026521599292755126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,8,balanced,0.05402666827042898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,16,power_law_1.2,0.22897920608520508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,16,power_law_1.2,0.041254401206970215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,8,balanced,0.05489066739877065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,16,power_law_1.2,0.31882879734039304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,8,balanced,0.05198400219281515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,16,power_law_1.2,0.04320639967918396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,16,power_law_1.2,0.37973120212554934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,8,balanced,0.05455466608206431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,16,power_law_1.2,0.04257279932498932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,16,power_law_1.2,0.5856448173522949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,8,balanced,0.053488001227378845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,16,power_law_1.2,0.04127359986305237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,16,power_law_1.2,0.8312064170837402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,8,balanced,0.05233600238958994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,16,power_law_1.2,0.04185599982738495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,8,balanced,0.05529066423575083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,16,power_law_1.2,1.1672320365905762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,16,power_law_1.2,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,8,balanced,0.05352533360322317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,16,power_law_1.2,0.04286080002784729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,16,power_law_1.2,2.794547271728516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,8,balanced,0.05435733497142792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,16,power_law_1.2,0.04056319892406464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,8,balanced,0.0609386662642161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,16,power_law_1.2,0.04459519982337952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,8,balanced,0.05946133534113566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,16,power_law_1.2,0.043680000305175784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.2,0.04746240079402923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,8,balanced,0.06452799836794536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.2,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,8,balanced,0.06725333134333293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.2,0.06631039977073669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,8,balanced,0.07089066505432129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.2,0.08429440259933471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.2,0.09850239753723145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.2,0.13068799972534179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,8,balanced,0.07877333462238312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.2,0.16897280216217042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.2,0.2322688102722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,8,balanced,0.0883626639842987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.2,0.2970880031585693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,16,power_law_1.2,0.3691391944885254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,8,balanced,0.10702932874361674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,16,power_law_1.2,0.5679743766784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,16,power_law_1.2,1.1480256080627442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,8,balanced,0.11602133512496948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,8,balanced,0.14655466874440512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,8,balanced,0.1882773240407308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,8,balanced,0.24949334065119425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,8,balanced,0.31691733996073407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,8,balanced,0.3868906497955322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,8,balanced,0.034287999073664345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,8,balanced,0.037018666664759316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,8,balanced,0.0390079990029335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,8,balanced,0.04363733530044556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,8,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,8,balanced,0.051125332713127136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,8,balanced,0.04889066517353058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,8,balanced,0.05128533144791921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,8,balanced,0.05068266888459524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,8,balanced,0.05091733237107595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,8,balanced,0.05006400247414907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,8,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,8,balanced,0.05305600166320801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,8,balanced,0.052613332867622375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,8,balanced,0.582751989364624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,8,balanced,0.05656533439954122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,8,balanced,0.05932266513506571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,8,balanced,0.05989866455396017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,8,balanced,0.06506133576234181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,8,power_law_1.01,0.047814399003982544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,8,balanced,0.06780266761779785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,8,power_law_1.01,0.04245119988918304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,8,power_law_1.01,0.040992000699043275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,8,balanced,0.08495466907819112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,8,power_law_1.01,0.045817598700523376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,8,power_law_1.01,0.04780800044536591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,8,power_law_1.01,0.05141760110855102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,8,balanced,0.09778666496276855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,8,power_law_1.01,0.05456640124320984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,8,power_law_1.01,0.05363839864730835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,8,power_law_1.01,0.05603839755058289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,8,balanced,0.13049599528312683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,8,power_law_1.01,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,8,power_law_1.01,0.05384960174560547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,8,power_law_1.01,0.0545087993144989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,8,balanced,0.14870400230089822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,8,power_law_1.01,0.054636800289154054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,8,power_law_1.01,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,8,power_law_1.01,0.061510401964187625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,8,power_law_1.01,0.06265599727630615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,8,power_law_1.01,0.04428159892559051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,8,balanced,0.2018079956372579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,8,power_law_1.01,0.06595839858055115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,8,power_law_1.01,0.03893119990825653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,8,power_law_1.01,0.07283200025558471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,8,power_law_1.01,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,8,power_law_1.01,0.08078719973564148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,8,power_law_1.01,0.041555199027061465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,8,power_law_1.01,0.09003520011901855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,8,power_law_1.01,0.04343039989471435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,8,power_law_1.01,0.10113919973373413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,8,power_law_1.01,0.04746879935264588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,8,power_law_1.01,0.1275712013244629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,8,balanced,0.25387734174728394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,8,power_law_1.01,0.04783360064029694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,8,power_law_1.01,0.158188796043396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,8,power_law_1.01,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,8,power_law_1.01,0.20554239749908448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,8,power_law_1.01,0.04835839867591858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,8,power_law_1.01,0.2693120002746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,8,power_law_1.01,0.050547200441360476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,8,power_law_1.01,0.3851392030715942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,8,power_law_1.01,0.05012480020523071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,8,power_law_1.01,0.49429121017456057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,8,power_law_1.01,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,8,power_law_1.01,0.5494719982147217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,8,power_law_1.01,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,8,balanced,0.35571734110514325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,8,power_law_1.01,0.8501248359680176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,8,power_law_1.01,0.0542527973651886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,8,power_law_1.01,0.05857920050621033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,8,power_law_1.01,1.8166271209716798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,8,power_law_1.01,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,8,power_law_1.01,0.06361600160598754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,8,balanced,1.0984426339467366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,8,power_law_1.01,0.07195519804954528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,8,power_law_1.01,0.08229119777679443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,8,power_law_1.01,0.10176639556884766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,8,power_law_1.01,0.11660159826278686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,8,power_law_1.01,0.14161920547485352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,8,power_law_1.01,0.18455040454864502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,8,balanced,0.4612319866816203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,8,power_law_1.01,0.2425663948059082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,8,power_law_1.01,0.3329024076461792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,8,power_law_1.01,0.46726398468017577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,8,power_law_1.01,0.6413951873779297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,8,power_law_1.01,0.7903488159179688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,8,power_law_1.01,1.2729984283447267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,8,power_law_1.01,2.3706687927246093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,8,balanced,0.5662986834843954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,8,balanced,0.8774346510569254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,8,balanced,1.7001546223958333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,8,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,8,power_law_1.2,0.04801279902458191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,8,power_law_1.2,0.037990400195121767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,8,power_law_1.2,0.04116480052471161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,8,power_law_1.2,0.039468801021575926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,8,power_law_1.2,0.04193919897079468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,8,power_law_1.2,0.043859198689460754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,8,power_law_1.2,0.04131839871406555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,8,power_law_1.2,0.046982398629188536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,8,power_law_1.2,0.043808001279830935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,8,power_law_1.2,0.05382400155067444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,8,power_law_1.2,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,8,power_law_1.2,0.051948797702789304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,8,power_law_1.2,0.04757120013237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,8,power_law_1.2,0.05466880202293396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,8,power_law_1.2,0.04852479994297028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,8,power_law_1.2,0.05616000294685364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,8,power_law_1.2,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,8,power_law_1.2,0.05621119737625122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,8,power_law_1.2,0.049670401215553286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,8,power_law_1.2,0.05437440276145935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,8,power_law_1.2,0.050400000810623166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,8,power_law_1.2,0.056595200300216676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,8,power_law_1.2,0.0491456001996994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,4,balanced,0.038058665891488395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,8,power_law_1.2,0.055199998617172244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,8,power_law_1.2,0.05343359708786011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,4,balanced,0.04205866654713949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,8,power_law_1.2,0.05934720039367676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,8,power_law_1.2,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,4,balanced,0.044666667779286705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,8,power_law_1.2,0.06162559986114502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,8,power_law_1.2,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,4,balanced,0.05115200082461039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,8,power_law_1.2,0.06475520133972168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,8,power_law_1.2,0.061414402723312375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,4,balanced,0.06339199841022491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,4,balanced,0.06304533282915752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,8,power_law_1.2,0.07011200189590454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,8,power_law_1.2,0.06614400148391723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,4,balanced,0.06506133576234181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,8,power_law_1.2,0.07377279996871948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,8,power_law_1.2,0.07563520073890687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,4,balanced,0.06374933322270711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,8,power_law_1.2,0.0800320029258728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,8,power_law_1.2,0.08215680122375488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,4,balanced,0.06205866734186808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,8,power_law_1.2,0.10060160160064698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,8,power_law_1.2,0.1013375997543335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,4,balanced,0.06125866870085398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,8,power_law_1.2,0.10398080348968505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,8,power_law_1.2,0.12372479438781739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,4,balanced,0.061018665631612144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,8,power_law_1.2,0.13890559673309327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,8,power_law_1.2,0.16743040084838867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,4,balanced,0.0644053320089976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,4,balanced,0.06327466666698456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,8,power_law_1.2,0.16353919506072997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,8,power_law_1.2,0.20003840923309327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,4,balanced,0.06505066653092702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,8,power_law_1.2,0.25167360305786135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,8,power_law_1.2,0.32283520698547363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,4,balanced,0.06821866830190022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,8,power_law_1.2,0.30702719688415525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,8,power_law_1.2,0.3731391906738281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,4,balanced,0.0711893339951833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,8,power_law_1.2,0.4099775791168213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,8,power_law_1.2,0.5310783863067627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,4,balanced,0.07131200035413106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,8,power_law_1.2,0.5967167854309082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,8,power_law_1.2,0.7677760124206543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,8,power_law_1.2,0.7726975917816162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,8,power_law_1.2,0.8483839988708496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,4,balanced,0.08171199758847554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,8,power_law_1.2,1.344985580444336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,8,power_law_1.2,1.4987135887145997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,4,balanced,0.08595200379689534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,8,power_law_1.2,2.2441280364990233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,8,power_law_1.2,3.4200897216796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,4,balanced,0.09724799791971843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,4,balanced,0.1067626674969991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,4,balanced,0.1361066699028015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,4,balanced,0.15500266353289285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,4,balanced,0.22038400173187256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,4,balanced,0.274725337823232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,4,balanced,0.38291200002034503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,4,balanced,0.03566399961709976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,4,balanced,0.03876800090074539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,4,balanced,0.040576001008351646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,4,balanced,0.04691733419895172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,4,balanced,0.057328000664711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,4,balanced,0.05871466795603434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,4,balanced,0.05643199880917867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,4,balanced,0.058448001742362976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,4,balanced,0.4950186808904012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,4,balanced,0.058730666836102806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,4,balanced,0.05685333410898844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,4,balanced,0.0592853327592214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,4,balanced,0.060821334520975746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,4,balanced,0.061205332477887474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,4,balanced,0.06090133388837179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,4,balanced,0.06508799890677135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,4,balanced,0.0676746666431427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,4,balanced,0.07409066458543141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,4,balanced,0.07945600152015686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,4,balanced,0.08719999591509502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,4,balanced,0.10794132947921753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,4,balanced,0.6059733231862386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,4,balanced,0.1220960021018982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,4,balanced,0.1655946671962738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,4,balanced,0.19269333283106485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,4,balanced,0.2718240022659302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,4,balanced,0.3455893198649089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,4,balanced,0.9294719696044922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,4,power_law_1.01,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,4,balanced,0.5020159880320231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,4,power_law_1.01,0.03922559916973114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,4,power_law_1.01,0.045510399341583255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,4,power_law_1.01,0.04023039937019348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,4,power_law_1.01,0.04403199851512909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,4,power_law_1.01,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,4,power_law_1.01,0.04320639967918396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,4,power_law_1.01,0.04589439928531647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,4,power_law_1.01,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,4,power_law_1.01,0.05464959740638733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,4,power_law_1.01,0.05172479748725891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,4,power_law_1.01,0.0542464017868042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,4,power_law_1.01,0.05973119735717773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,4,power_law_1.01,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,4,power_law_1.01,0.061561602354049685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,4,power_law_1.01,0.05548160076141358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,4,power_law_1.01,0.06284160017967225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,4,power_law_1.01,0.05678079724311828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,4,power_law_1.01,0.0639360010623932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,4,power_law_1.01,0.060012799501419065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,4,balanced,0.6498506863911947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,4,power_law_1.01,0.0616703987121582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,4,power_law_1.01,0.060819202661514284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,4,power_law_1.01,0.06428800225257873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,4,power_law_1.01,0.06087679862976074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,4,power_law_1.01,0.06255360245704651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,4,power_law_1.01,0.06424319744110107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,4,power_law_1.01,0.06677119731903076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,4,power_law_1.01,0.06496000289916992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,4,power_law_1.01,0.06993920207023621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,4,power_law_1.01,0.0679423987865448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,4,power_law_1.01,0.07159039974212647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,4,power_law_1.01,0.07073280215263367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,4,power_law_1.01,0.08802559971809387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,4,power_law_1.01,0.07327359914779663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,4,power_law_1.01,0.09998720288276672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,4,power_law_1.01,0.07854719758033753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,4,power_law_1.01,0.08764799833297729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,4,power_law_1.01,0.12474240064620971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,4,power_law_1.01,0.0945792019367218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,4,power_law_1.01,0.13935359716415405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,4,power_law_1.01,0.11073280572891235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,4,power_law_1.01,0.2077183961868286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,4,balanced,0.8018506368001302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,4,power_law_1.01,0.12081279754638671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,4,power_law_1.01,0.22092161178588868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,4,power_law_1.01,0.16606080532073975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,4,power_law_1.01,0.3083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,4,power_law_1.01,0.19676159620285033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,4,power_law_1.01,0.4041088104248047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,4,power_law_1.01,0.2823551893234253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,4,power_law_1.01,0.5784832000732422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,4,power_law_1.01,0.33902080059051515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,4,power_law_1.01,0.7766464233398438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,4,power_law_1.01,0.46460161209106443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,4,power_law_1.01,0.9645503997802735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,4,power_law_1.01,0.6750400066375732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,4,power_law_1.01,1.6119680404663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,4,power_law_1.01,0.7894976139068604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,4,power_law_1.01,3.0060800552368163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,4,power_law_1.01,1.2315584182739259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,4,power_law_1.01,2.236319923400879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,4,balanced,1.8242026964823406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,4,balanced,1.2413067022959392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,4,balanced,2.4451573689778647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,4,power_law_1.2,0.043244799971580504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,4,power_law_1.2,0.03814400136470795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,4,power_law_1.2,0.03944959938526153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,4,power_law_1.2,0.046137601137161255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,4,power_law_1.2,0.043910399079322815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,4,power_law_1.2,0.04249599874019623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,4,power_law_1.2,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,2,balanced,0.04212800165017446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,4,power_law_1.2,0.0547327995300293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,4,power_law_1.2,0.04249599874019623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,2,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,4,power_law_1.2,0.05622400045394897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,4,power_law_1.2,0.04760960042476654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,2,balanced,0.049098665515581764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,4,power_law_1.2,0.056620800495147706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,4,power_law_1.2,0.052691197395324706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,2,balanced,0.05817066629727682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,4,power_law_1.2,0.055820798873901366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,4,power_law_1.2,0.06058880090713501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,2,balanced,0.07358400026957194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,2,balanced,0.07381333410739899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,4,power_law_1.2,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,4,power_law_1.2,0.05792639851570129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,2,balanced,0.07204266885916392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,4,power_law_1.2,0.06417279839515685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,4,power_law_1.2,0.05921919941902161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,2,balanced,0.07145066559314728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,4,power_law_1.2,0.060729598999023436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,4,power_law_1.2,0.06179839968681335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,2,balanced,0.07387199997901917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,4,power_law_1.2,0.06356480121612548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,4,power_law_1.2,0.06123520135879516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,2,balanced,0.07301333546638489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,4,power_law_1.2,0.06330879926681518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,4,power_law_1.2,0.061561602354049685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,2,balanced,0.07467199862003326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,4,power_law_1.2,0.0650111973285675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,4,power_law_1.2,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,2,balanced,0.07587199906508128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,4,power_law_1.2,0.06538239717483521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,4,power_law_1.2,0.07103999853134155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,2,balanced,0.07760533193747203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,4,power_law_1.2,0.06902400255203248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,4,power_law_1.2,0.07674880027770996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,4,power_law_1.2,0.08858879804611205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,2,balanced,0.08002666632334392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,4,power_law_1.2,0.10460159778594971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,2,balanced,0.08851200342178345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,4,power_law_1.2,0.12451839447021484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,2,balanced,0.08806932965914409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,4,power_law_1.2,0.1447167992591858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,2,balanced,0.09435199697812398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,4,power_law_1.2,0.18577280044555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,4,power_law_1.2,0.2340991973876953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,2,balanced,0.10418666402498881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,4,power_law_1.2,0.3191551923751831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,4,power_law_1.2,0.4539968013763428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,2,balanced,0.11363200346628825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,4,power_law_1.2,0.6712192058563232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,4,power_law_1.2,0.8724608421325684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,2,balanced,0.13007466991742453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,4,power_law_1.2,1.1466239929199218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,2,balanced,0.1455359955628713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,4,power_law_1.2,1.964998435974121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,4,power_law_1.2,3.698566436767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,2,balanced,0.204202671845754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,2,balanced,0.24084800481796265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,2,balanced,0.3468000094095866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,2,balanced,0.03994133323431015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,2,balanced,0.043178667624791466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,2,balanced,0.046709333856900535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,2,balanced,0.4418506622314453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,2,balanced,0.05956799785296122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,2,balanced,0.07127999762694041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,2,balanced,0.07302933434645335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,2,balanced,0.07281066477298737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,2,balanced,0.07349333167076111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,2,balanced,0.07376533250013988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,2,balanced,0.07453333338101704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,2,balanced,0.07528000076611836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,2,balanced,0.07726400097211202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,2,balanced,0.07934399942557017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,2,balanced,0.6350080172220866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,2,balanced,0.07860266665617625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,2,balanced,0.08769599596659343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,2,balanced,0.08669867118199666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,2,balanced,0.09175466497739156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,2,balanced,0.10827733079592387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,2,balanced,0.1253866652647654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,2,balanced,0.14632532993952432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,2,balanced,0.812725305557251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,2,balanced,0.16881599028905234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,2,balanced,0.23750933011372885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,2,balanced,0.28866666555404663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,2,balanced,1.0083146890004475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,2,balanced,0.41738665103912354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,2,balanced,0.5380586783091227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,2,balanced,1.5550133387247722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,2,balanced,0.7950987021128336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,2,balanced,1.0395999749501545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,2,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,2,power_law_1.01,0.04307839870452881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,2,power_law_1.01,0.04693120121955872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,4,power_law_1.2,0.0725823998451233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,2,power_law_1.01,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,4,power_law_1.2,0.07456640005111695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,2,power_law_1.01,0.0568448007106781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,2,power_law_1.01,0.0709119975566864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,4,power_law_1.2,0.08060160279273987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,2,power_law_1.01,0.07100160121917724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,4,power_law_1.2,0.08947839736938476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,2,power_law_1.01,0.07186560034751892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,4,power_law_1.2,0.09801599979400635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,2,power_law_1.01,0.07389439940452576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,4,power_law_1.2,0.11670399904251098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,2,power_law_1.01,0.07371519804000855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,4,power_law_1.2,0.1344704031944275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,2,power_law_1.01,0.07640320062637329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,4,power_law_1.2,0.17696640491485596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,2,power_law_1.01,0.07705600261688232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,4,power_law_1.2,0.21774721145629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,2,power_law_1.01,0.07904639840126038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,4,power_law_1.2,0.2940608024597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,2,balanced,1.2923413117726643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,2,power_law_1.01,0.08315520286560059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,4,power_law_1.2,0.43306879997253417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,2,power_law_1.01,0.0871616005897522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,4,power_law_1.2,0.4841728210449219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,2,power_law_1.01,0.08841599822044373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,4,power_law_1.2,0.7186880111694336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,2,power_law_1.01,0.09690240025520325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,2,power_law_1.01,0.10744960308074951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,4,power_law_1.2,0.9860032081604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,2,power_law_1.01,0.12008960247039795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,4,power_law_1.2,1.580134391784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,2,power_law_1.01,0.1419263958930969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,2,power_law_1.01,0.1627392053604126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,4,power_law_1.2,3.4802112579345703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,2,balanced,3.0385545094807944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,2,power_law_1.01,0.22186880111694335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,2,power_law_1.01,0.2878783941268921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,2,power_law_1.01,0.38020479679107666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,2,power_law_1.01,0.5192704200744629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,2,power_law_1.01,0.7054080009460449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,2,power_law_1.01,0.9243136405944824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,2,power_law_1.01,1.0958080291748047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,2,power_law_1.01,1.828108787536621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,2,power_law_1.01,3.3858814239501953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,2,balanced,2.015082677205404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,2,balanced,3.9761972427368164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,2,power_law_1.01,0.043033599853515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,2,power_law_1.2,0.0440064013004303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,2,power_law_1.01,0.04261119961738587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,2,power_law_1.2,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,2,power_law_1.01,0.044563201069831845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,2,power_law_1.2,0.04438399970531463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,2,power_law_1.01,0.05061759948730469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,2,power_law_1.2,0.052236801385879515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,2,power_law_1.2,0.04343680143356323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,2,power_law_1.01,0.05553920269012451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,2,power_law_1.2,0.05504639744758606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,2,power_law_1.2,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,2,power_law_1.01,0.06772480010986329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,2,power_law_1.2,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,2,power_law_1.2,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,2,power_law_1.2,0.07027199864387512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,2,power_law_1.01,0.07071999907493591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,2,power_law_1.2,0.05097600221633911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,2,power_law_1.01,0.070278400182724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,2,power_law_1.2,0.07440639734268188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,2,power_law_1.2,0.05591679811477661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,2,power_law_1.01,0.07309439778327942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,2,power_law_1.2,0.07349119782447815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,2,power_law_1.2,0.06813439726829529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,2,power_law_1.01,0.07218559980392455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,2,power_law_1.2,0.07440000176429748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,2,power_law_1.2,0.0680895984172821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,2,power_law_1.01,0.07475200295448303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,2,power_law_1.2,0.07128959894180298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,2,power_law_1.2,0.0755840003490448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,2,power_law_1.01,0.07600640058517456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,2,power_law_1.2,0.07315840125083924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,2,power_law_1.2,0.07886719703674316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,2,power_law_1.01,0.07677440047264099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,2,power_law_1.2,0.07328640222549439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,2,power_law_1.2,0.0790336012840271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,2,power_law_1.01,0.07797759771347046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,2,power_law_1.2,0.07480319738388061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,2,power_law_1.2,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,2,power_law_1.01,0.08576639890670776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,2,power_law_1.2,0.07565439939498901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,2,power_law_1.2,0.0896511971950531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,2,power_law_1.01,0.08986240029335021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,2,power_law_1.2,0.0769216001033783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,2,power_law_1.2,0.09127680063247681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,2,power_law_1.01,0.09589759707450866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,2,power_law_1.2,0.07754240036010743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,2,power_law_1.2,0.09691519737243652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,2,power_law_1.01,0.11607680320739747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,2,power_law_1.2,0.08521599769592285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,2,power_law_1.2,0.10959999561309815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,2,power_law_1.01,0.13037439584732055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,2,power_law_1.2,0.08880000114440918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,2,power_law_1.2,0.12113280296325683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,2,power_law_1.01,0.15544960498809815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,2,power_law_1.2,0.0987775981426239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,2,power_law_1.2,0.1438655972480774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,2,power_law_1.01,0.1964735984802246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,2,power_law_1.2,0.11850240230560302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,2,power_law_1.2,0.17794560194015502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,2,power_law_1.01,0.26031999588012694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,2,power_law_1.2,0.1334015965461731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,2,power_law_1.2,0.2414016008377075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,2,power_law_1.01,0.31933441162109377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,2,power_law_1.2,0.16286079883575438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,2,power_law_1.2,0.29468159675598143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,2,power_law_1.2,0.19509119987487794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,2,power_law_1.01,0.4562816143035889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,2,power_law_1.2,0.40560641288757326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,2,power_law_1.2,0.2580032110214233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,2,power_law_1.01,0.5974207878112793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,2,power_law_1.2,0.5446656227111817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,2,power_law_1.2,0.3429951906204224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,2,power_law_1.01,0.825004768371582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,2,power_law_1.2,0.7113152027130127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,2,power_law_1.2,0.46620798110961914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,2,power_law_1.01,1.131161594390869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,2,power_law_1.2,0.9172927856445312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,2,power_law_1.2,0.6031040191650391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,2,power_law_1.01,1.4343168258666992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,2,power_law_1.2,1.1058496475219726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,2,power_law_1.2,0.9009984016418457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,2,power_law_1.01,2.169164848327637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,2,power_law_1.2,2.018367958068848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,2,power_law_1.2,1.1729663848876952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,2,power_law_1.01,4.305574417114258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,2,power_law_1.2,3.971974563598633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,2,power_law_1.2,1.4290431976318358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,2,power_law_1.2,2.434432029724121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,2,power_law_1.2,4.896294403076172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,32,balanced,0.06140799820423126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,32,balanced,0.0653706689675649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,32,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,32,balanced,0.07897066573301952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,32,balanced,0.09795733292897542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,32,balanced,0.11727999647458394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,32,balanced,0.1070240040620168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,32,balanced,0.11314133803049724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,32,balanced,0.11062933007876079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,32,balanced,0.10213866829872131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,32,balanced,0.10441600282986958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,32,balanced,0.10833600163459778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,32,balanced,0.11027733484903972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,32,balanced,0.10406399766604106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,32,balanced,0.11077866951624553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,32,balanced,0.11413333813349406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,32,balanced,0.10820266604423523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,32,balanced,0.11997866630554199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,32,balanced,0.11645866433779399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,32,balanced,0.12313600381215413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,32,balanced,0.13065600395202637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,32,balanced,0.18145066499710083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,32,balanced,0.19010132551193237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,32,balanced,0.2392746607462565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,32,balanced,0.3015519976615906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,32,balanced,0.36986132462819415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,32,balanced,0.054101333022117615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,32,balanced,0.05561600128809611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,32,balanced,0.48499735196431476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,32,balanced,0.057850668827692665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,32,balanced,0.053632001082102455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,32,balanced,0.06733866532643636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,32,balanced,0.05518400172392527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,32,balanced,0.07341333230336507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,32,balanced,0.055071999629338585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,32,balanced,0.08208000163237254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,32,balanced,0.061893333991368614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,32,balanced,0.025519999365011852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,32,balanced,0.0806879997253418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,32,balanced,0.0729066679875056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,32,balanced,0.026341333985328674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,32,balanced,0.07784000039100647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,32,balanced,0.08065066734949748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,32,balanced,0.025834667185942333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,32,balanced,0.028149334092934925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,32,balanced,0.07716266810894012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,32,balanced,0.07950933277606964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,32,balanced,0.03070399910211563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,32,balanced,0.0839359958966573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,32,balanced,0.07894933223724365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,32,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,32,balanced,0.08556800087292989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,32,balanced,0.08086400230725606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,32,balanced,0.055018668373425804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,32,balanced,0.08216533561547597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,32,balanced,0.08229333162307739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,32,balanced,0.05653866628805796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,32,balanced,0.08959999680519104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,32,balanced,0.08185600241025288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,32,balanced,0.05698133508364359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,32,balanced,0.08427733182907104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,32,balanced,0.08332799871762593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,32,balanced,0.06308799982070923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,32,balanced,0.0830026666323344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,32,balanced,0.06107733150323232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,32,balanced,0.08700799942016602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,32,balanced,0.08761066198348999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,32,balanced,0.5470346609751383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,32,balanced,0.05704000095526377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,32,balanced,0.09011200070381165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,32,balanced,0.08851733803749084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,32,balanced,0.05416533350944519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,32,balanced,0.09155199925104777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,32,balanced,0.0906986693541209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,32,balanced,0.050517335534095764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,32,balanced,0.09242133299509685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,32,balanced,0.07313066720962524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,32,balanced,0.0979253351688385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,32,balanced,0.07214400172233582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,32,balanced,0.09950400392214458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,32,balanced,0.10124799609184265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,32,balanced,0.06809066732724507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,32,balanced,0.10612799723943074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,32,balanced,0.11162133018175761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,32,balanced,0.07152533531188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,32,balanced,0.1225440005461375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,32,balanced,0.0899839997291565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,32,balanced,0.1195199986298879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,32,balanced,0.11357866724332173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,32,balanced,0.13876799742380777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,32,balanced,0.16430399815241495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,32,balanced,0.13498133420944214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,32,balanced,0.18152532974878946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,32,balanced,0.1726613243420919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,32,balanced,0.18158400058746338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,32,balanced,0.21024533112843832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,32,balanced,0.2270080049832662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,32,balanced,0.83133331934611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,32,balanced,0.22792534033457437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,32,balanced,0.32108267148335773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,32,balanced,0.275818665822347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,32,balanced,0.2845653295516968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,32,balanced,0.41287465890248615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,32,balanced,0.34988800684611004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,32,balanced,0.3643519878387451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,32,balanced,0.6106826861699423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,32,balanced,0.46754666169484455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,32,balanced,0.478000005086263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,32,balanced,0.8064853350321451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,32,balanced,0.6098666588465372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,32,balanced,0.5541386604309082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,32,balanced,1.632522741953532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,32,balanced,0.995797316233317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,32,balanced,0.7223626772562662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,32,balanced,0.8443413575490316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,32,balanced,1.5772266387939453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,32,balanced,1.1158933639526367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,32,balanced,3.113807996114095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,32,balanced,1.6501973470052083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,32,balanced,2.196239948272705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,32,power_law_1.01,0.1015936017036438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,32,power_law_1.01,0.09711359739303589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,32,power_law_1.01,0.08240640163421631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,32,power_law_1.01,0.08518400192260742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,32,power_law_1.01,0.09601280093193054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,32,power_law_1.01,0.10646400451660157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,32,power_law_1.01,0.10520319938659668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,32,power_law_1.01,0.1048192024230957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,32,power_law_1.01,0.10602879524230957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,32,power_law_1.01,0.10347520112991333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,32,power_law_1.01,0.10390399694442749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,32,power_law_1.01,0.10046720504760742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,32,power_law_1.01,0.10392320156097412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,32,power_law_1.01,0.10497920513153076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,32,power_law_1.01,0.10516480207443238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,32,power_law_1.01,0.1066431999206543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,32,power_law_1.01,0.11207679510116578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,32,power_law_1.01,0.1327296018600464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,32,power_law_1.01,0.14456959962844848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,32,power_law_1.01,0.16763520240783691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,32,power_law_1.01,0.18494720458984376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,32,power_law_1.01,0.2361664056777954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,32,power_law_1.01,0.2825344085693359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,32,power_law_1.01,0.37360639572143556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,32,power_law_1.01,0.5142975807189941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,32,power_law_1.01,0.7062016010284424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,32,power_law_1.01,0.8507072448730468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,32,power_law_1.01,1.0749119758605956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,32,power_law_1.01,1.6052799224853516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,32,power_law_1.01,3.6218048095703126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,32,power_law_1.01,0.07704319953918456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,32,power_law_1.01,0.07112320065498352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,32,power_law_1.01,0.058182400465011594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,32,power_law_1.01,0.06840320229530335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,32,power_law_1.01,0.0716863989830017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,32,power_law_1.01,0.08156800270080566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,32,power_law_1.01,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,32,power_law_1.01,0.07958400249481201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,32,power_law_1.01,0.07941120266914367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,32,power_law_1.01,0.08202239871025085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,32,power_law_1.01,0.07950080037117005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,32,power_law_1.01,0.08350719809532166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,32,power_law_1.01,0.08006399869918823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,32,power_law_1.01,0.08140799999237061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,32,power_law_1.01,0.05805439949035644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,32,power_law_1.01,0.08335360288619995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,32,power_law_1.01,0.06910719871520996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,32,power_law_1.01,0.08528000116348267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,32,power_law_1.01,0.07153919935226441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,32,power_law_1.01,0.08785279989242553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,32,power_law_1.01,0.07490559816360473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,32,power_law_1.01,0.08887040019035339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,32,power_law_1.01,0.07614719867706299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,32,power_law_1.01,0.09681919813156128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,32,power_law_1.01,0.07856000065803528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,32,power_law_1.01,0.10446079969406127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,32,power_law_1.01,0.07722240090370178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,32,power_law_1.01,0.1259071946144104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,32,power_law_1.01,0.024409599602222443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,32,power_law_1.01,0.07830399870872498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,32,power_law_1.01,0.1492416024208069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,32,power_law_1.01,0.02536959946155548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,32,power_law_1.01,0.07797120213508606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,32,power_law_1.01,0.16999679803848267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,32,power_law_1.01,0.025183999538421632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,32,power_law_1.01,0.07908480167388916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,32,power_law_1.01,0.21752960681915284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,32,power_law_1.01,0.024806399643421174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,32,power_law_1.01,0.0811456024646759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,32,power_law_1.01,0.2736128091812134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,32,power_law_1.01,0.029311999678611755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,32,power_law_1.01,0.08204159736633301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,32,power_law_1.01,0.03758080005645752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,32,power_law_1.01,0.3770240068435669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,32,power_law_1.01,0.08564479947090149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,32,power_law_1.01,0.05496960282325745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,32,power_law_1.01,0.43978238105773926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,32,power_law_1.01,0.08594560027122497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,32,power_law_1.01,0.054655998945236206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,32,power_law_1.01,0.6539584159851074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,32,power_law_1.01,0.05482879877090454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,32,power_law_1.01,0.09182080030441284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,32,power_law_1.01,0.887564754486084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,32,power_law_1.01,0.0616703987121582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,32,power_law_1.01,0.10431360006332398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,32,power_law_1.01,1.04333438873291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,32,power_law_1.01,0.06170880198478699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.01,0.11730560064315795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,32,power_law_1.01,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.01,0.13511040210723876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,32,power_law_1.01,1.7535615921020509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,32,power_law_1.01,0.06053119897842407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.01,0.14689919948577881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,32,power_law_1.01,3.3951297760009767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,32,power_law_1.01,0.058143997192382814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.01,0.20482559204101564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,32,power_law_1.01,0.05674880146980286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.01,0.22818560600280763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,32,power_law_1.01,0.057004797458648684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.01,0.29198079109191893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,32,power_law_1.01,0.06180480122566223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.01,0.3939199924468994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,32,power_law_1.01,0.06567040085792542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.01,0.5330560207366943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.01,0.06780800223350525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.01,0.6910655975341797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.01,0.07900800108909607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,32,power_law_1.01,0.8792384147644043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.01,0.0906175971031189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,32,power_law_1.01,1.359769630432129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.01,0.12083840370178223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.01,0.1515391945838928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,32,power_law_1.01,2.7486719131469726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.01,0.19759999513626098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.01,0.2642431974411011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.01,0.36355199813842776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.01,0.49509758949279786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,32,power_law_1.01,0.5715328216552734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,32,power_law_1.01,0.9912639617919922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,32,power_law_1.01,1.943212890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,32,power_law_1.2,0.10099200010299683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,32,power_law_1.2,0.08145920038223267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,32,power_law_1.2,0.07719039916992188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,32,power_law_1.2,0.07722240090370178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,32,power_law_1.2,0.09472000002861022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,32,power_law_1.2,0.10443520545959473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,32,power_law_1.2,0.10081919431686401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,32,power_law_1.2,0.10533759593963624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,32,power_law_1.2,0.10376960039138794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,32,power_law_1.2,0.10517120361328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,32,power_law_1.2,0.10247679948806762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,32,power_law_1.2,0.10414079427719117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,32,power_law_1.2,0.10386559963226319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,32,power_law_1.2,0.10466560125350952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,32,power_law_1.2,0.10958080291748047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,32,power_law_1.2,0.11327999830245972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,32,power_law_1.2,0.1260159969329834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,32,power_law_1.2,0.14099199771881105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,32,power_law_1.2,0.15550719499588012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,32,power_law_1.2,0.17923840284347534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,32,power_law_1.2,0.21478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,32,power_law_1.2,0.25932159423828127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,32,power_law_1.2,0.31585919857025146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,32,power_law_1.2,0.42690558433532716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,32,power_law_1.2,0.5529471874237061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,32,power_law_1.2,0.8286911964416503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,32,power_law_1.2,1.1269760131835938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,32,power_law_1.2,1.357215976715088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,32,power_law_1.2,2.4443647384643556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,32,power_law_1.2,5.518387222290039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,32,power_law_1.2,0.07875840067863464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,32,power_law_1.2,0.06839680075645446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,32,power_law_1.2,0.06223359704017639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,32,power_law_1.2,0.0712768018245697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,32,power_law_1.2,0.07372159957885742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,32,power_law_1.2,0.07487360239028931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,32,power_law_1.2,0.07857919931411743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,32,power_law_1.2,0.07677440047264099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,32,power_law_1.2,0.07962239980697632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,32,power_law_1.2,0.08145279884338379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,32,power_law_1.2,0.08197759985923767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,32,power_law_1.2,0.08395519852638245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,32,power_law_1.2,0.08330240249633789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,32,power_law_1.2,0.07642239928245545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,32,power_law_1.2,0.08546559810638428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,32,power_law_1.2,0.06696959733963012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,32,power_law_1.2,0.09045760035514831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,32,power_law_1.2,0.06030719876289368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,32,power_law_1.2,0.06550400257110596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,32,power_law_1.2,0.09256960153579712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,32,power_law_1.2,0.07171199917793274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,32,power_law_1.2,0.10392320156097412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,32,power_law_1.2,0.07646080255508422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,32,power_law_1.2,0.11650559902191163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,32,power_law_1.2,0.0774399995803833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,32,power_law_1.2,0.12807040214538573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,32,power_law_1.2,0.07646719813346863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,32,power_law_1.2,0.15768959522247314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,32,power_law_1.2,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,32,power_law_1.2,0.18430720567703246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,32,power_law_1.2,0.07694720029830933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,32,power_law_1.2,0.024857600033283234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,32,power_law_1.2,0.2441472053527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,32,power_law_1.2,0.07996159791946411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,32,power_law_1.2,0.024025599658489227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,32,power_law_1.2,0.3161855936050415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,32,power_law_1.2,0.07980800271034241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,32,power_law_1.2,0.024659200012683867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,32,power_law_1.2,0.43471360206604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,32,power_law_1.2,0.08247039914131164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,32,power_law_1.2,0.02648960053920746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,32,power_law_1.2,0.5644032001495362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,32,power_law_1.2,0.08120319843292237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,32,power_law_1.2,0.028672000765800475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,32,power_law_1.2,0.9152959823608399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,32,power_law_1.2,0.08536319732666016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,32,power_law_1.2,0.03741439878940582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,32,power_law_1.2,0.9821951866149903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,32,power_law_1.2,0.0913856029510498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,32,power_law_1.2,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,32,power_law_1.2,1.3120320320129395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,32,power_law_1.2,0.0984063982963562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,32,power_law_1.2,0.053855997323989865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,32,power_law_1.2,0.11232000589370728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,32,power_law_1.2,2.364102363586426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,32,power_law_1.2,0.052716797590255736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.2,0.11872639656066894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,32,power_law_1.2,0.06193280220031738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,32,power_law_1.2,4.5276542663574215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.2,0.14613120555877684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,32,power_law_1.2,0.0636672019958496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.2,0.15949440002441406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,32,power_law_1.2,0.06111360192298889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.2,0.2056191921234131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,32,power_law_1.2,0.061939197778701785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.2,0.27187199592590333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,32,power_law_1.2,0.057580798864364624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.2,0.37633280754089354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,32,power_law_1.2,0.054022401571273804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.2,0.49373440742492675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,32,power_law_1.2,0.056428802013397214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.2,0.7209152221679688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,32,power_law_1.2,0.05551360249519348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.2,0.9868927955627441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,32,power_law_1.2,0.06785280108451844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.2,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,32,power_law_1.2,1.1020992279052735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.2,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,32,power_law_1.2,1.7485952377319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.2,0.10133119821548461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,32,power_law_1.2,4.0635009765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.2,0.1282304048538208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.2,0.15418879985809325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.2,0.2206399917602539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.2,0.2790080070495605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.2,0.4647103786468506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.2,0.6278207778930665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,32,power_law_1.2,0.7016575813293457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,32,power_law_1.2,1.306387233734131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,32,power_law_1.2,2.4366783142089843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,16,balanced,0.05438933273156484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,16,balanced,0.0580320010582606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,16,balanced,0.060991997520128884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,16,balanced,0.07533866663773854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,16,balanced,0.09448533256848653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,16,balanced,0.10522666573524475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,16,balanced,0.10063999891281128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,16,balanced,0.1006773312886556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,16,balanced,0.1081119974454244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,16,balanced,0.09968533118565877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,16,balanced,0.10140267014503479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,16,balanced,0.0981119970480601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,16,balanced,0.09798399607340495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,16,balanced,0.10176533460617065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,16,balanced,0.10805333654085796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,16,balanced,0.10728533069292705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,16,balanced,0.10829866925875346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,16,balanced,0.11289067069689433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,16,balanced,0.11606933673222859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,16,balanced,0.1256480018297831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,16,balanced,0.13616533080736795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,16,balanced,0.18200000127156576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,16,balanced,0.19735999902089438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,16,balanced,0.04990933338801066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,16,balanced,0.05092266698678335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,16,balanced,0.053210665782292686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,16,balanced,0.06283733248710632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,16,balanced,0.06840000053246816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,16,balanced,0.2497119903564453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,16,balanced,0.07786133388678233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,16,balanced,0.07789333164691925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,16,balanced,0.08215466638406117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,16,balanced,0.08379200100898743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,16,balanced,0.0788266658782959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,16,balanced,0.08191466828187306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,16,balanced,0.0836853285630544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,16,balanced,0.08156266808509827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,16,balanced,0.08701866865158081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,16,balanced,0.3176906704902649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,16,balanced,0.08877866466840108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,16,balanced,0.0900320013364156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,16,balanced,0.05251200000445048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,16,balanced,0.09355733791987102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,16,balanced,0.053120002150535583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,16,balanced,0.05754133562246958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,16,balanced,0.10377599795659383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,16,balanced,0.07019199927647908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,16,balanced,0.075162669022878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,16,balanced,0.11038933197657268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,16,balanced,0.08595732847849528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,16,balanced,0.08546666304270427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,16,balanced,0.08190933366616567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,16,balanced,0.12838400403658548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,16,balanced,0.08640533685684204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,16,balanced,0.08369599779446919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,16,balanced,0.3999893267949422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,16,balanced,0.14451199769973755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,16,balanced,0.08346133430798848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,16,balanced,0.023919999599456787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,16,balanced,0.0249493345618248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,16,balanced,0.08356266220410664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,16,balanced,0.025045332809289295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,16,balanced,0.08359466989835103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,16,balanced,0.027098665634791057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,16,balanced,0.1913386583328247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,16,balanced,0.08570133646329244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,16,balanced,0.029077333708604176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,16,balanced,0.09172800183296204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,16,balanced,0.03215466688076655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,16,balanced,0.03765333443880081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,16,balanced,0.09107733766237895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,16,balanced,0.036874666810035706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,16,balanced,0.09366400043169658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,16,balanced,0.2198986609776815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,16,balanced,0.0455626646677653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,16,balanced,0.05563200016816457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,16,balanced,0.09847467144330342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,16,balanced,0.05575466652711233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,16,balanced,0.10320533315340678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,16,balanced,0.06155199805895487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,16,balanced,0.06352533400058746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,16,balanced,0.29450666904449463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,16,balanced,0.11851732929547627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,16,balanced,0.06043200194835663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,16,balanced,0.07259733478228252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,16,balanced,0.5168266693751017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,16,balanced,0.07507200042406718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,16,balanced,0.1295253336429596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,16,balanced,0.06984533369541168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,16,balanced,0.37085334459940594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,16,balanced,0.07966400186220805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,16,balanced,0.1728000044822693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,16,balanced,0.09075733025868733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,16,balanced,0.18824533621470133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,16,balanced,0.10942400495211284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,16,balanced,0.12417599558830261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,16,balanced,0.5048693418502808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,16,balanced,0.2508853276570638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,16,balanced,0.15494400262832642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,16,balanced,0.5960693359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,16,balanced,0.1836693286895752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,16,balanced,0.31697599093119305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,16,balanced,0.26417599121729535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,16,balanced,0.6556320190429688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,16,balanced,0.41333866119384766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,16,balanced,0.3285599946975708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,16,balanced,0.7866026560465494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,16,balanced,0.5436160167058309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,16,balanced,0.47257598241170246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,16,balanced,0.9140480359395345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,16,balanced,0.6253600120544434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,16,balanced,0.6377226511637369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,16,balanced,1.2190186977386475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,16,balanced,0.7548213005065918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,16,balanced,0.9822399616241455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,16,balanced,1.1973386605580647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,16,balanced,1.7989439964294434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,16,balanced,2.400442600250244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,16,balanced,1.9233867327372234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,16,balanced,2.345242659250895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,16,power_law_1.01,0.09207680225372314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,16,power_law_1.01,0.08549759984016418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,16,power_law_1.01,0.06969599723815918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,16,power_law_1.01,0.07544320225715637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,16,power_law_1.01,0.08417919874191285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,16,power_law_1.01,0.09351680278778077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,16,power_law_1.01,0.09655680060386658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,16,power_law_1.01,0.10029439926147461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,16,power_law_1.01,0.09834880232810975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,16,power_law_1.01,0.09847679734230042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,16,power_law_1.01,0.09770879745483399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,16,power_law_1.01,0.1005247950553894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,16,power_law_1.01,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,16,power_law_1.01,0.0975488007068634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,16,power_law_1.01,0.07335039973258972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,16,power_law_1.01,0.10028159618377686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,16,power_law_1.01,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,16,power_law_1.01,0.10753920078277587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,16,power_law_1.01,0.06622080206871032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,16,power_law_1.01,0.10435839891433715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,16,power_law_1.01,0.06878719925880432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,16,power_law_1.01,0.11078399419784546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,16,power_law_1.01,0.12023040056228637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,16,power_law_1.01,0.07685760259628296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,16,power_law_1.01,0.13024640083312988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,16,power_law_1.01,0.08004480004310607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,16,power_law_1.01,0.15859199762344361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,16,power_law_1.01,0.07824000120162963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,16,power_law_1.01,0.17832959890365602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,16,power_law_1.01,0.08000640273094177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,16,power_law_1.01,0.19783040285110473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,16,power_law_1.01,0.081004798412323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,16,power_law_1.01,0.08165760040283203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,16,power_law_1.01,0.27866239547729493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,16,power_law_1.01,0.08021119832992554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,16,power_law_1.01,0.07970560193061829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,16,power_law_1.01,0.3744447946548462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,16,power_law_1.01,0.08198400139808655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,16,power_law_1.01,0.06250240206718445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,16,power_law_1.01,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,16,power_law_1.01,0.415558385848999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,16,power_law_1.01,0.06707839965820313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,16,power_law_1.01,0.08536959886550903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,16,power_law_1.01,0.6472000122070313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,16,power_law_1.01,0.07111039757728577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,16,power_law_1.01,0.0902079999446869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,16,power_law_1.01,0.7585663795471191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,16,power_law_1.01,0.08046720027923585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,16,power_law_1.01,0.09185280203819275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,16,power_law_1.01,0.982579231262207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,16,power_law_1.01,0.08213120102882385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,16,power_law_1.01,0.09591680169105529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,16,power_law_1.01,1.4762687683105469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,16,power_law_1.01,0.08276479840278625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,16,power_law_1.01,0.11052800416946411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,16,power_law_1.01,0.08326399922370911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,16,power_law_1.01,3.126515197753906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,16,power_law_1.01,0.12350720167160034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,16,power_law_1.01,0.08448640108108521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,16,power_law_1.01,0.15015039443969727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,16,power_law_1.01,0.0851967990398407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,16,power_law_1.01,0.18059519529342652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,16,power_law_1.01,0.0865343987941742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,16,power_law_1.01,0.22239999771118163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,16,power_law_1.01,0.08759679794311523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,16,power_law_1.01,0.2559231996536255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,16,power_law_1.01,0.08798080086708068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,16,power_law_1.01,0.023852799832820893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,16,power_law_1.01,0.3720576047897339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,16,power_law_1.01,0.09268479943275451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,16,power_law_1.01,0.025484800338745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,16,power_law_1.01,0.48698878288269043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,16,power_law_1.01,0.09325439929962158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,16,power_law_1.01,0.022969600558280946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,16,power_law_1.01,0.7287360191345215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,16,power_law_1.01,0.09849600195884704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,16,power_law_1.01,0.024908800423145295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,16,power_law_1.01,0.8470463752746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,16,power_law_1.01,0.11144959926605225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,16,power_law_1.01,0.02536959946155548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.01,0.12247680425643921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,16,power_law_1.01,1.0903231620788574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,16,power_law_1.01,0.029465600848197937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.01,0.14126720428466796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,16,power_law_1.01,0.03797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,16,power_law_1.01,1.6414144515991211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.01,0.16353919506072997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,16,power_law_1.01,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,16,power_law_1.01,3.279391860961914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.01,0.20748159885406495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,16,power_law_1.01,0.036735999584198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.01,0.2483135938644409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,16,power_law_1.01,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.01,0.3135040044784546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,16,power_law_1.01,0.055174398422241214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.01,0.39450879096984864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,16,power_law_1.01,0.054771202802658084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,16,power_law_1.01,0.0606656014919281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.01,0.6152256011962891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,16,power_law_1.01,0.06419839859008789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.01,0.7744832038879395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,16,power_law_1.01,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,16,power_law_1.01,0.9207936286926269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,16,power_law_1.01,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,16,power_law_1.01,1.3814720153808593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,16,power_law_1.01,0.06280320286750793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,16,power_law_1.01,2.927142333984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,16,power_law_1.01,0.061791998147964475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.01,0.06768640279769897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.01,0.08261119723320007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.01,0.09130240082740784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.01,0.11127680540084839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.01,0.1356160044670105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.01,0.18403840065002441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.01,0.22581119537353517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.01,0.31445119380950926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.01,0.3999743938446045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,16,power_law_1.01,0.5092735767364502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,16,power_law_1.01,0.8114239692687988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,16,power_law_1.01,1.5158080101013183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,16,power_law_1.2,0.09160959720611572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,16,power_law_1.2,0.0692351996898651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,16,power_law_1.2,0.06951040029525757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,16,power_law_1.2,0.07789440155029297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,16,power_law_1.2,0.0813759982585907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,16,power_law_1.2,0.09904000163078308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,16,power_law_1.2,0.09342719912528992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,16,power_law_1.2,0.09728000164031983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,16,power_law_1.2,0.09589759707450866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,16,power_law_1.2,0.09607040286064147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,16,power_law_1.2,0.10087679624557495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,16,power_law_1.2,0.10113279819488526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,16,power_law_1.2,0.10183039903640748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,16,power_law_1.2,0.0738752007484436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,16,power_law_1.2,0.10389120578765869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,16,power_law_1.2,0.06001920104026794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,16,power_law_1.2,0.10618879795074462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,16,power_law_1.2,0.057062399387359616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,16,power_law_1.2,0.10899200439453124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,16,power_law_1.2,0.06570240259170532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,16,power_law_1.2,0.11570559740066529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,16,power_law_1.2,0.06629760265350342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,16,power_law_1.2,0.12975360155105592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,16,power_law_1.2,0.07471359968185425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,16,power_law_1.2,0.14971519708633424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,16,power_law_1.2,0.07626879811286927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,16,power_law_1.2,0.16635520458221437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,16,power_law_1.2,0.07713919878005981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,16,power_law_1.2,0.1965824007987976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,16,power_law_1.2,0.07994880080223084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,16,power_law_1.2,0.08009600043296813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,16,power_law_1.2,0.24119679927825927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,16,power_law_1.2,0.08229759931564332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,16,power_law_1.2,0.29332480430603025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,16,power_law_1.2,0.08297600150108338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,16,power_law_1.2,0.4063551902770996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,16,power_law_1.2,0.08293120265007019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,16,power_law_1.2,0.49677438735961915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,16,power_law_1.2,0.08320000171661376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,16,power_law_1.2,0.7775487899780273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,16,power_law_1.2,0.09285759925842285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,16,power_law_1.2,0.08142079710960388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,16,power_law_1.2,0.9015168190002442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,16,power_law_1.2,0.09420160055160523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,16,power_law_1.2,0.06074879765510559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,16,power_law_1.2,1.3041279792785645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,16,power_law_1.2,0.09918720126152039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,16,power_law_1.2,0.06110079884529114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,16,power_law_1.2,2.1524736404418947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,16,power_law_1.2,0.11067520380020142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,16,power_law_1.2,0.06750720143318176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,16,power_law_1.2,0.12513920068740844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,16,power_law_1.2,0.06926720142364502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,16,power_law_1.2,4.174745559692383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,16,power_law_1.2,0.08329600095748901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,16,power_law_1.2,0.1564288020133972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,16,power_law_1.2,0.0827072024345398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,16,power_law_1.2,0.17121280431747438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,16,power_law_1.2,0.08213120102882385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,16,power_law_1.2,0.23412480354309081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,16,power_law_1.2,0.08067839741706848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,16,power_law_1.2,0.2936448097229004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,16,power_law_1.2,0.08259199857711792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,16,power_law_1.2,0.4080192089080811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,16,power_law_1.2,0.08262400031089782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,16,power_law_1.2,0.4947391986846924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,16,power_law_1.2,0.08562560081481933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,16,power_law_1.2,0.8202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,16,power_law_1.2,0.023379200696945192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,16,power_law_1.2,0.08734719753265381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,16,power_law_1.2,1.0285696029663085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,16,power_law_1.2,0.021990400552749634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,16,power_law_1.2,0.08544639945030212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,16,power_law_1.2,1.2528767585754395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,16,power_law_1.2,0.022579200565814972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,16,power_law_1.2,0.09326080083847046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,16,power_law_1.2,0.023852799832820893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,16,power_law_1.2,2.1050880432128904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,16,power_law_1.2,0.09483519792556763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,16,power_law_1.2,0.026150399446487428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,16,power_law_1.2,0.10113919973373413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,16,power_law_1.2,4.2172096252441404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,16,power_law_1.2,0.029600000381469725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,16,power_law_1.2,0.11208959817886352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,16,power_law_1.2,0.03688960075378418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.2,0.13023359775543214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,16,power_law_1.2,0.03726080060005188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.2,0.142739200592041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.2,0.17382400035858153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,16,power_law_1.2,0.035155200958251955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.2,0.21742720603942872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,16,power_law_1.2,0.05367680191993714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.2,0.2756416082382202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,16,power_law_1.2,0.054067200422286986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,16,power_law_1.2,0.05529599785804749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.2,0.382259202003479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,16,power_law_1.2,0.06257280111312866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.2,0.4363840103149414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,16,power_law_1.2,0.06231679916381836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.2,0.6420351982116699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,16,power_law_1.2,0.05989120006561279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.2,0.9047167778015137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,16,power_law_1.2,0.06099200248718262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,16,power_law_1.2,1.1255999565124513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,16,power_law_1.2,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,16,power_law_1.2,1.7755647659301759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,16,power_law_1.2,0.05926399827003479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.2,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,16,power_law_1.2,3.7251518249511717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.2,0.0855679988861084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.2,0.09185280203819275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.2,0.11974400281906128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.2,0.14492160081863403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.2,0.18761600255966188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.2,0.2453696012496948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.2,0.35376639366149903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.2,0.46416640281677246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,16,power_law_1.2,0.5719168186187744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,16,power_law_1.2,0.9603008270263672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,16,power_law_1.2,1.8769920349121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,8,balanced,0.05401599903901418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,8,balanced,0.05760000149408976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,8,balanced,0.06118933359781901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,8,balanced,0.07737066845099132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,8,balanced,0.09497066338857015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,8,balanced,0.10014933347702026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,8,balanced,0.10205333431561787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,8,balanced,0.09937600294748943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,8,balanced,0.10053867101669312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,8,balanced,0.09644266963005066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,8,balanced,0.09704533219337463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,8,balanced,0.09797333677609761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,8,balanced,0.050213331977526345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,8,balanced,0.09611200292905171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,8,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,8,balanced,0.097871998945872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,8,balanced,0.055104002356529236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,8,balanced,0.06307200094064076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,8,balanced,0.10459733009338379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,8,balanced,0.07663999994595845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,8,balanced,0.10611200332641602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,8,balanced,0.08657067020734151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,8,balanced,0.08493866523106892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,8,balanced,0.10909333825111389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,8,balanced,0.08361599842707317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,8,balanced,0.11783466736475627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,8,balanced,0.08706133564313252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,8,balanced,0.08559466401735942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,8,balanced,0.12040000160535176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,8,balanced,0.08643200000127156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,8,balanced,0.08762666583061218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,8,balanced,0.13521599769592285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,8,balanced,0.0874826709429423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,8,balanced,0.09171199798583984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,8,balanced,0.09711999694506328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,8,balanced,0.14727999766667685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,8,balanced,0.09711999694506328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,8,balanced,0.10053867101669312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,8,balanced,0.1980959971745809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,8,balanced,0.11132267117500305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,8,balanced,0.12090133627255757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,8,balanced,0.14076266686121622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,8,balanced,0.22054932514826456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,8,balanced,0.15665599703788757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,8,power_law_1.01,0.07760639786720276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,8,power_law_1.01,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,8,balanced,0.2152000069618225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,8,balanced,0.28147733211517334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,8,power_law_1.01,0.0673088014125824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,8,power_law_1.01,0.07839360237121581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,8,power_law_1.01,0.08391039967536926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,8,balanced,0.25082133213679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,8,power_law_1.01,0.09296000003814697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,8,power_law_1.01,0.09689599871635438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,8,power_law_1.01,0.09715840220451355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,8,power_law_1.01,0.09798399806022644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,8,balanced,0.35396265983581543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,8,power_law_1.01,0.09693440198898315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,8,balanced,0.3380906581878662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,8,power_law_1.01,0.10165760517120362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,8,power_law_1.01,0.10000640153884888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,8,power_law_1.01,0.09907839894294738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,8,power_law_1.01,0.10406399965286255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,8,power_law_1.01,0.10829440355300904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,8,power_law_1.01,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,8,balanced,0.43059198061625165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,8,power_law_1.01,0.10915839672088623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,8,power_law_1.01,0.07022079825401306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,8,power_law_1.01,0.11486079692840576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,8,power_law_1.01,0.05644800066947937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,8,power_law_1.01,0.12345600128173828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,8,balanced,0.46990398565928143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,8,power_law_1.01,0.06478719711303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,8,power_law_1.01,0.1313920021057129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,8,power_law_1.01,0.07006080150604248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,8,power_law_1.01,0.16698880195617677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,8,power_law_1.01,0.07809919714927674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,8,power_law_1.01,0.18684799671173097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,8,power_law_1.01,0.08290560245513916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,8,power_law_1.01,0.2367232084274292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,8,power_law_1.01,0.0833791971206665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,8,power_law_1.01,0.2695552110671997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,8,balanced,0.598202665646871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,8,power_law_1.01,0.08331519961357117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,8,power_law_1.01,0.3630592107772827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,8,power_law_1.01,0.08478720188140869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,8,power_law_1.01,0.4243008136749268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,8,power_law_1.01,0.08595200181007386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,8,power_law_1.01,0.6335103988647461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,8,power_law_1.01,0.08538879752159119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,8,power_law_1.01,0.7942527770996094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,8,power_law_1.01,0.08693119883537292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,8,power_law_1.01,1.0637503623962403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,8,power_law_1.01,0.09009280204772949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,8,balanced,0.6031306584676107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,8,power_law_1.01,1.5489791870117187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,8,power_law_1.01,0.09738879799842834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,8,power_law_1.01,0.09978240132331848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,8,power_law_1.01,3.017740821838379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,8,power_law_1.01,0.10451840162277222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,8,balanced,0.7747573057810465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,8,power_law_1.01,0.11890560388565063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,8,power_law_1.01,0.1314687967300415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,8,power_law_1.01,0.16694400310516358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,8,power_law_1.01,0.19007999897003175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,8,power_law_1.01,0.23185279369354247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,8,power_law_1.01,0.3014656066894531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,8,power_law_1.01,0.41479039192199707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,8,power_law_1.01,0.4953023910522461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,8,power_law_1.01,0.768556785583496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,8,power_law_1.01,0.962495994567871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,8,balanced,0.9376533031463623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,8,balanced,0.7138826847076416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,8,power_law_1.01,1.245145606994629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,8,power_law_1.01,1.8616512298583985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,8,power_law_1.01,3.6962558746337892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,8,balanced,1.453386624654134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,8,balanced,1.0858666896820068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,8,balanced,2.8722079594930015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,8,balanced,2.1382880210876465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,8,power_law_1.2,0.0752128005027771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,8,power_law_1.2,0.06512640118598938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,8,power_law_1.2,0.0662015974521637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,8,power_law_1.2,0.07481600046157837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,8,power_law_1.2,0.08289920091629029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,8,power_law_1.2,0.09283199906349182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,8,power_law_1.2,0.09723520278930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,8,power_law_1.2,0.09811840057373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,8,power_law_1.2,0.10026880502700805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,8,power_law_1.2,0.09844480156898498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,8,power_law_1.2,0.0998080015182495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,8,power_law_1.2,0.10087679624557495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,8,power_law_1.2,0.10095360279083251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,8,power_law_1.2,0.1046720027923584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,8,power_law_1.2,0.11117440462112427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,8,power_law_1.2,0.11360640525817871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,8,power_law_1.2,0.11535359621047973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,8,power_law_1.2,0.12887040376663209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,8,power_law_1.2,0.0650111973285675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,8,power_law_1.2,0.15416959524154664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,8,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,8,power_law_1.2,0.16503039598464966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,8,power_law_1.2,0.05919359922409058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,8,power_law_1.2,0.18695679903030396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,8,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,8,power_law_1.2,0.23785600662231446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,8,power_law_1.2,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,8,power_law_1.2,0.30076799392700193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,8,power_law_1.2,0.07914239764213563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,8,power_law_1.2,0.08262400031089782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,8,power_law_1.2,0.41199359893798826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,8,power_law_1.2,0.08221439719200134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,8,power_law_1.2,0.5503488063812256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,8,power_law_1.2,0.0833728015422821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,8,power_law_1.2,0.6875711917877197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,8,power_law_1.2,0.08661760091781616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,8,power_law_1.2,0.8670975685119628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,8,power_law_1.2,0.08736000061035157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,8,power_law_1.2,1.213913631439209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,8,power_law_1.2,0.08748160004615783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,8,power_law_1.2,1.7042623519897462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,8,power_law_1.2,0.08837760090827942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,8,power_law_1.2,0.09599999785423279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,8,power_law_1.2,4.206022262573242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,8,power_law_1.2,0.10032000541687011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,8,power_law_1.2,0.10436480045318604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,8,power_law_1.2,0.10836479663848878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,8,power_law_1.2,0.1255295991897583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,8,power_law_1.2,0.1423424005508423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,8,power_law_1.2,0.1776576042175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,8,power_law_1.2,0.19203840494155883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,8,power_law_1.2,0.2609407901763916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,8,power_law_1.2,0.30471038818359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,8,power_law_1.2,0.44173440933227537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,8,power_law_1.2,0.5790656089782715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,8,power_law_1.2,0.8231679916381835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,8,power_law_1.2,1.1713791847229005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,8,power_law_1.2,1.3781184196472167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,8,power_law_1.2,2.054220771789551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,8,power_law_1.2,4.554137420654297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,4,balanced,0.05657066901524862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,4,balanced,0.06185600161552429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,4,balanced,0.0644053320089976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,4,balanced,0.08177599807580312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,4,balanced,0.105621337890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,4,balanced,0.11781332890192668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,4,balanced,0.11528000235557556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,4,balanced,0.11572800079981486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,4,balanced,0.05128000179926554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,4,balanced,0.1129919985930125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,4,balanced,0.05507733424504598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,4,balanced,0.11238400141398112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,4,balanced,0.060826669136683144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,4,balanced,0.11269866426785786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,4,balanced,0.06782933572928111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,4,balanced,0.0958079993724823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,4,balanced,0.11220266421635945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,4,balanced,0.09807466467221577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,4,balanced,0.11290132999420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,4,balanced,0.10029866298039754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,4,balanced,0.11530133088429768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,4,balanced,0.10204266508420308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,4,balanced,0.118559996287028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,4,balanced,0.10143466790517171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,4,balanced,0.09662399689356486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,4,balanced,0.12191466490427653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,4,balanced,0.09995733698209126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,4,balanced,0.127018670241038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,4,balanced,0.10140800476074219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,4,balanced,0.09930666287740071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,4,balanced,0.13643200198809305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,4,balanced,0.10366400082906087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,4,balanced,0.1450826625029246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,4,balanced,0.10999466975529988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,4,balanced,0.11215466260910034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,4,balanced,0.16361600160598755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,4,balanced,0.12033599615097046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,4,balanced,0.13365333278973898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,4,balanced,0.17857066790262857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,4,balanced,0.14191466569900513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,4,balanced,0.167738676071167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,4,balanced,0.2507466673851013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,4,balanced,0.1918933391571045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,4,balanced,0.2840213378270467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,4,balanced,0.26889065901438397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,4,balanced,0.3905706803003947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,4,balanced,0.31471999486287433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,4,balanced,0.4431946674982707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,4,balanced,0.4979039827982585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,4,balanced,0.5685919920603434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,4,balanced,0.6520373423894247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,4,balanced,0.7812799612681071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,4,power_law_1.01,0.0673088014125824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,4,balanced,0.8625173568725586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,4,power_law_1.01,0.07735679745674133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,4,power_law_1.01,0.06840320229530335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,4,power_law_1.01,0.0812991976737976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,4,balanced,1.0269920031229656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,4,power_law_1.01,0.08926079869270324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,4,power_law_1.01,0.10547840595245361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,4,power_law_1.01,0.10595840215682983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,4,power_law_1.01,0.10740480422973633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,4,power_law_1.01,0.10798079967498779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,4,power_law_1.01,0.10814080238342286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,4,power_law_1.01,0.10886399745941162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,4,power_law_1.01,0.11084159612655639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,4,power_law_1.01,0.1130560040473938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,4,power_law_1.01,0.11477119922637939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,4,balanced,1.0310453573862712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,4,power_law_1.01,0.12314879894256592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,4,power_law_1.01,0.1227455973625183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,4,balanced,1.2465120156606038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,4,power_law_1.01,0.1295680046081543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,4,power_law_1.01,0.14379520416259767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,4,power_law_1.01,0.06105599999427795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,4,power_law_1.01,0.1704192042350769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,4,power_law_1.01,0.06123520135879516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,4,power_law_1.01,0.19900799989700318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,4,power_law_1.01,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,4,power_law_1.01,0.21561601161956787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,4,power_law_1.01,0.06902400255203248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,4,power_law_1.01,0.27971200942993163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,4,power_law_1.01,0.0770687997341156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,4,power_law_1.01,0.3205120086669922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,4,power_law_1.01,0.08821759819984436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,4,power_law_1.01,0.4530943870544434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,4,power_law_1.01,0.09455999732017517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,4,power_law_1.01,0.5801536083221436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,4,power_law_1.01,0.09367679953575134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,4,power_law_1.01,0.8089152336120605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,4,power_law_1.01,0.0956928014755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,4,power_law_1.01,0.993286418914795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,4,power_law_1.01,0.09638400077819824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,4,power_law_1.01,1.3409728050231933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,4,power_law_1.01,0.09818239808082581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,4,power_law_1.01,1.963430404663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,4,power_law_1.01,0.10210560560226441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,4,balanced,1.6157600084940593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,4,power_law_1.01,0.10283520221710205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,4,power_law_1.01,4.159609603881836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,4,power_law_1.01,0.10559359788894654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,4,balanced,1.9558827082316081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,4,power_law_1.01,0.11096320152282715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,4,power_law_1.01,0.11509759426116943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,4,power_law_1.01,0.12145919799804687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,4,power_law_1.01,0.13642879724502563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,4,power_law_1.01,0.1568511962890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,4,power_law_1.01,0.19077759981155396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,4,power_law_1.01,0.22683520317077638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,4,power_law_1.01,0.29002881050109863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,4,power_law_1.01,0.38782079219818116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,4,power_law_1.01,0.5013823986053467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,4,power_law_1.01,0.6368576049804687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,4,power_law_1.01,0.9195199966430664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,4,power_law_1.01,1.1878975868225097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,4,power_law_1.01,1.4182720184326172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,4,power_law_1.01,2.4274879455566407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,4,power_law_1.01,4.60682258605957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,4,balanced,3.1751200358072915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,4,balanced,3.8633012771606445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,4,power_law_1.2,0.06754559874534607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,4,power_law_1.2,0.06764159798622131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,4,power_law_1.2,0.07095680236816407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,4,power_law_1.2,0.07582079768180847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,4,power_law_1.2,0.08709759712219238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,4,power_law_1.2,0.10359679460525513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,4,power_law_1.2,0.10739840269088745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,4,power_law_1.2,0.11034879684448243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,4,power_law_1.2,0.10781439542770385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,4,power_law_1.2,0.10998400449752807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,4,power_law_1.2,0.11025279760360718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,4,power_law_1.2,0.11288959980010986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,4,power_law_1.2,0.11496319770812988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,4,power_law_1.2,0.11868799924850464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,4,power_law_1.2,0.12702720165252684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,4,power_law_1.2,0.12651519775390624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,4,power_law_1.2,0.1345471978187561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,4,power_law_1.2,0.1455615997314453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,4,power_law_1.2,0.18026880025863648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,4,power_law_1.2,0.20556159019470216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,4,power_law_1.2,0.06056320071220398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,4,power_law_1.2,0.23432319164276122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,4,power_law_1.2,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,4,power_law_1.2,0.28236799240112304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,4,power_law_1.2,0.059385597705841064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,4,power_law_1.2,0.35177600383758545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,4,power_law_1.2,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,4,power_law_1.2,0.48478078842163086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,4,power_law_1.2,0.07639039754867553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,2,balanced,0.06070399781068166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,4,power_law_1.2,0.7227007865905761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,4,power_law_1.2,0.09234560132026673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,2,balanced,0.06505600114663442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,4,power_law_1.2,0.8975999832153321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,2,balanced,0.054383998115857445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,2,balanced,0.07787199815114339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,4,power_law_1.2,0.09835519790649414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,2,balanced,0.05902933577696482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,2,balanced,0.0953546663125356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,4,power_law_1.2,1.0553919792175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,4,power_law_1.2,0.09895679950714112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,2,balanced,0.06489600241184235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,2,balanced,0.13101866841316223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,4,power_law_1.2,0.09695360064506531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,2,balanced,0.08067200084527333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,4,power_law_1.2,1.450425624847412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,2,balanced,0.16765334208806357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,2,balanced,0.11123200257619222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,2,balanced,0.15827199816703796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,4,power_law_1.2,0.09784960150718688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,4,power_law_1.2,2.6786048889160154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,2,balanced,0.12774399916330972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,2,balanced,0.15837333599726358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,4,power_law_1.2,0.09994239807128906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,2,balanced,0.12944533427556357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,2,balanced,0.15552000204722086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,4,power_law_1.2,4.844377517700195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,4,power_law_1.2,0.09932799935340882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,2,balanced,0.13141866525014242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,2,balanced,0.1591093341509501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,4,power_law_1.2,0.10490880012512208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,2,balanced,0.13306666413942972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,2,balanced,0.156741331020991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,2,balanced,0.13277332981427512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,2,balanced,0.1575093368689219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,4,power_law_1.2,0.1044160008430481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,2,balanced,0.13285332918167114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,2,balanced,0.15902933478355408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,4,power_law_1.2,0.11463040113449097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,2,balanced,0.13595733046531677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,2,balanced,0.16084800163904825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,4,power_law_1.2,0.11960959434509277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,2,balanced,0.13531733552614847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,2,balanced,0.1660373310248057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,4,power_law_1.2,0.1260480046272278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,2,balanced,0.14138666788736978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,2,balanced,0.16793066263198853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,4,power_law_1.2,0.14142719507217408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,2,balanced,0.14628266294797262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,2,balanced,0.17249600092569986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,4,power_law_1.2,0.17144960165023804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,2,balanced,0.15075199802716574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,4,power_law_1.2,0.19800959825515746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,2,balanced,0.18888533115386963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,2,balanced,0.15965333580970764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,4,power_law_1.2,0.22979838848114015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,2,balanced,0.2014133334159851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,4,power_law_1.2,0.31515519618988036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,2,balanced,0.18058133125305176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,4,power_law_1.2,0.3805824041366577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,2,balanced,0.22913066546122232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,2,balanced,0.19555733601252237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,4,power_law_1.2,0.5291135787963868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,4,power_law_1.2,0.7137343883514404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,2,balanced,0.23109867175420126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,2,balanced,0.25702399015426636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,4,power_law_1.2,1.0101632118225097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,4,power_law_1.2,1.248960018157959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,2,balanced,0.26410667101542157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,2,balanced,0.37491198380788165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,4,power_law_1.2,1.6634239196777343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,4,power_law_1.2,2.602707290649414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,2,balanced,0.3882346550623576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,2,balanced,0.4248480002085368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,4,power_law_1.2,5.174176025390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,2,balanced,0.4604959885279338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,2,balanced,0.5956213474273682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,2,balanced,0.6559093395868937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,2,balanced,0.7618559996287028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,2,balanced,0.8472320238749186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,2,balanced,1.026799996693929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,2,balanced,1.1744159857432048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,2,balanced,1.3638505935668945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,2,balanced,1.551194667816162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,2,balanced,1.6453173955281575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,2,balanced,1.8775733311971028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,2,balanced,2.5851945877075195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,2,balanced,2.982581456502279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,2,power_law_1.01,0.06481279730796814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,2,power_law_1.01,0.06798080205917359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,2,power_law_1.01,0.07583360075950622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,2,power_law_1.01,0.08881919980049133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,2,power_law_1.01,0.09990400075912476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,2,power_law_1.01,0.12572159767150878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,2,power_law_1.01,0.1313215970993042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,2,power_law_1.01,0.1306496024131775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,2,power_law_1.01,0.13169280290603638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,2,power_law_1.01,0.13696639537811278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,2,power_law_1.01,0.1336575984954834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,2,power_law_1.01,0.13631999492645264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,2,power_law_1.01,0.14083199501037597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,2,power_law_1.01,0.14825600385665894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,2,power_law_1.01,0.1508095979690552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,2,power_law_1.01,0.15371520519256593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,2,power_law_1.01,0.17820800542831422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,2,power_law_1.01,0.19544960260391236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,2,power_law_1.01,0.060063999891281125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,2,power_law_1.01,0.06295679807662964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,2,power_law_1.01,0.21023359298706054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,2,power_law_1.01,0.06415359973907471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,2,power_law_1.01,0.24618239402770997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,2,power_law_1.01,0.07491840124130249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,2,power_law_1.01,0.2856640100479126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,2,power_law_1.01,0.08763520121574402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,2,power_law_1.01,0.37781760692596433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,2,power_law_1.01,0.10794880390167236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,2,power_law_1.01,0.45759358406066897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,2,balanced,5.143882751464844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,2,power_law_1.01,0.11635199785232545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,2,power_law_1.01,0.6543488025665283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,2,balanced,5.934431711832683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,2,power_law_1.01,0.11608320474624634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,2,power_law_1.01,0.7901887893676758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,2,power_law_1.01,0.11987839937210083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,2,power_law_1.01,1.1353919982910157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,2,power_law_1.01,0.1195520043373108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,2,power_law_1.01,1.4880767822265626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,2,power_law_1.01,0.12421120405197143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,2,power_law_1.01,1.8417984008789063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,2,power_law_1.01,0.12687360048294066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,2,power_law_1.01,0.12981760501861572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,2,power_law_1.01,2.88088321685791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,2,power_law_1.01,0.1306496024131775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,2,power_law_1.01,5.775110244750977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,2,power_law_1.01,0.14071040153503417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,2,power_law_1.01,0.14600319862365724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,2,power_law_1.01,0.15650559663772584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,2,power_law_1.01,0.1865023970603943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,2,power_law_1.01,0.20531840324401857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,2,power_law_1.01,0.24780800342559814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,2,power_law_1.01,0.29839999675750734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,2,power_law_1.01,0.3954047918319702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,2,power_law_1.01,0.513644790649414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,2,power_law_1.01,0.6876480102539062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,2,power_law_1.01,0.8623744010925293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,2,power_law_1.01,1.3383168220520019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,2,power_law_1.01,1.6058687210083007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,2,power_law_1.01,2.049888038635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,2,power_law_1.01,3.4701313018798827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,2,power_law_1.01,6.3166656494140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,2,power_law_1.2,0.06098560094833374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,2,power_law_1.2,0.06606720089912414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,2,power_law_1.2,0.056857597827911374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,2,power_law_1.2,0.061913597583770755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,2,power_law_1.2,0.06498559713363647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,2,power_law_1.2,0.07602559924125671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,2,power_law_1.2,0.07276160120964051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,2,power_law_1.2,0.08814719915390015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,2,power_law_1.2,0.08529919981956482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,2,power_law_1.2,0.1010815978050232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,2,power_law_1.2,0.10928000211715698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,2,power_law_1.2,0.12126079797744752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,2,power_law_1.2,0.11356799602508545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,2,power_law_1.2,0.1339903950691223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,2,power_law_1.2,0.11495039463043213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,2,power_law_1.2,0.12925440073013306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,64,balanced,0.08564800024032593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,2,power_law_1.2,0.11923840045928955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,2,power_law_1.2,0.13191039562225343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,64,balanced,0.08998933434486389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,64,balanced,0.08839466174443562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,2,power_law_1.2,0.1386304020881653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,2,power_law_1.2,0.12035839557647705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,64,balanced,0.08913600444793701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,2,power_law_1.2,0.12376960515975952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,2,power_law_1.2,0.14406399726867675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,64,balanced,0.10439466436704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,2,power_law_1.2,0.1282304048538208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,64,balanced,0.12760000427563986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,64,balanced,0.17619200547536215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,2,power_law_1.2,0.13181439638137818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,64,balanced,0.16937599579493204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,2,power_law_1.2,0.1366719961166382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,64,balanced,0.17084266742070517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,2,power_law_1.2,0.14869760274887084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,64,balanced,0.18000000715255737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,2,power_law_1.2,0.1540992021560669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,64,balanced,0.1863306760787964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,2,power_law_1.2,0.16352640390396117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,64,balanced,0.17883199453353882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,2,power_law_1.2,0.18587520122528076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,64,balanced,0.18219733238220215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,2,power_law_1.2,0.2137279987335205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,64,balanced,0.1834239959716797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,2,power_law_1.2,0.2661184072494507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,64,balanced,0.1856000026067098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,2,power_law_1.2,0.2946752071380615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,64,balanced,0.17973866065343222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,2,power_law_1.2,0.42029438018798826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,64,balanced,0.18979734182357788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,2,power_law_1.2,0.5164159774780274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,64,balanced,0.18752533197402954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,2,power_law_1.2,0.6813055992126464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,2,power_law_1.2,0.9587264060974121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,64,balanced,0.20028799772262573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,2,power_law_1.2,1.371724796295166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,64,balanced,0.2104746699333191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,2,power_law_1.2,1.7358783721923827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,2,power_law_1.2,2.1739648818969726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,64,balanced,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,64,balanced,0.22016533215840658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,64,balanced,0.0662613312403361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,2,power_law_1.2,3.663948822021484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,64,balanced,0.06480533381303151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,64,balanced,0.06956266860167186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,64,balanced,0.07096533477306366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,2,power_law_1.2,6.891238403320313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,64,balanced,0.2421440084775289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,64,balanced,0.09696533282597859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,64,balanced,0.11682666341463725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,64,balanced,0.1146453320980072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,64,balanced,0.12142933408419292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,64,balanced,0.2674613396326701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,64,balanced,0.11808000008265178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,64,balanced,0.11803199847539265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,64,balanced,0.11868799726168315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,64,balanced,0.1253653367360433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,64,balanced,0.12250666817029317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,64,balanced,0.31994666655858356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,64,balanced,0.12961600224177042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,64,balanced,0.13029332955678305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,64,balanced,0.14807466665903726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,64,balanced,0.1478559970855713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,64,balanced,0.4276426633199056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,64,balanced,0.15753066539764404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,64,balanced,0.17347200711568198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,64,balanced,0.19485332568486533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,64,balanced,0.5030080080032349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,64,balanced,0.2363306681315104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,64,balanced,0.2759573260943095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,64,balanced,0.35889601707458496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,64,balanced,0.6478933493296305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,64,balanced,0.46457068125406903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,64,balanced,0.6165279944737753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,64,balanced,0.8618666330973307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,64,balanced,0.7983946800231934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,64,balanced,1.2564266522725422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,64,balanced,1.0177546342213948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,2,power_law_1.2,0.15149439573287965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,2,power_law_1.2,0.14555519819259644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,64,balanced,0.06025066475073496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,2,power_law_1.2,0.1616320013999939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,64,balanced,0.062368000547091164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,2,power_law_1.2,0.16109440326690674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,64,balanced,0.0620000014702479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,64,balanced,0.06592000027497609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,2,power_law_1.2,0.16224000453948975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,64,balanced,0.07063466807206471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,2,power_law_1.2,0.17982720136642455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,64,balanced,0.0960586667060852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,2,power_law_1.2,0.19927680492401123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,64,balanced,0.11547199885050456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,64,balanced,0.12038933237393697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,2,power_law_1.2,0.21517438888549806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,64,balanced,1.5441865921020508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,64,balanced,0.12046933174133301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,2,power_law_1.2,0.2519871950149536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,64,balanced,0.11642666657765706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,2,power_law_1.2,0.30437119007110597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,64,balanced,0.11689066886901855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,2,power_law_1.2,0.3946624040603638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,64,balanced,0.11475732922554016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,64,balanced,0.12100799878438313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,2,power_law_1.2,0.4871488094329834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,64,balanced,0.11805867155392964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,2,power_law_1.2,0.6245952129364014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,64,balanced,0.12853333353996277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,2,power_law_1.2,0.890719985961914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,64,balanced,0.1237440009911855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,2,power_law_1.2,1.239027214050293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,64,balanced,0.12796800335248312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,2,power_law_1.2,1.6077951431274413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,64,balanced,0.13662399848302206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,2,power_law_1.2,1.9836223602294922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,64,balanced,0.14751999576886496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,2,power_law_1.2,3.1520383834838865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,64,balanced,2.3063573837280273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,64,balanced,0.157642662525177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,2,power_law_1.2,6.03889274597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,64,balanced,0.1695093313852946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,64,balanced,0.19790933529535928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,64,balanced,0.22232532501220703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,64,balanced,0.2829173405965169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,64,balanced,0.3900800148646037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,64,balanced,2.9649813969930015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,64,balanced,0.48757867018381756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,64,balanced,0.6413653294245402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,64,balanced,0.8038933277130127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,64,balanced,1.222271998723348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,64,power_law_1.01,0.1626304030418396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,64,power_law_1.01,0.1719040036201477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,64,power_law_1.01,0.1128767967224121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,64,power_law_1.01,0.17310719490051268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,64,power_law_1.01,0.11157759428024291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,64,power_law_1.01,0.11075839996337891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,64,power_law_1.01,0.11137280464172364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,64,power_law_1.01,0.12305279970169067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,64,power_law_1.01,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,64,power_law_1.01,0.14380799531936644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,64,power_law_1.01,0.08863999843597412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,64,power_law_1.01,0.14064639806747437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,64,power_law_1.01,0.09553279876708984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,64,power_law_1.01,0.16220799684524537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,64,power_law_1.01,0.09978240132331848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,64,power_law_1.01,0.16781439781188964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,64,power_law_1.01,0.10968960523605346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,64,power_law_1.01,0.17023359537124633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,64,power_law_1.01,0.10929919481277466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,64,power_law_1.01,0.10709120035171509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,64,power_law_1.01,0.16594560146331788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,64,balanced,2.3078400293986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,64,power_law_1.01,0.11783039569854736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,64,power_law_1.01,0.16800639629364014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,64,power_law_1.01,0.11573120355606079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,64,power_law_1.01,0.17066240310668945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,64,power_law_1.01,0.15909119844436645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,64,power_law_1.01,0.11947519779205322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,64,power_law_1.01,0.1677183985710144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,64,power_law_1.01,0.12289279699325562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,64,power_law_1.01,0.12300800085067749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,64,power_law_1.01,0.17712639570236205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,64,power_law_1.01,0.13222399950027466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,64,power_law_1.01,0.16906880140304564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,64,power_law_1.01,0.12912000417709352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,64,power_law_1.01,0.1801344037055969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,64,power_law_1.01,0.2030719995498657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,64,power_law_1.01,0.14198399782180787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,64,power_law_1.01,0.22386560440063477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,64,power_law_1.01,0.15498239994049073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,64,power_law_1.01,0.18730239868164061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,64,power_law_1.01,0.24722559452056886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,64,power_law_1.01,0.20955519676208495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,64,power_law_1.01,0.31018240451812745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,64,power_law_1.01,0.2650752067565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,64,power_law_1.01,0.3497024059295654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,64,power_law_1.01,0.11272959709167481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,64,power_law_1.01,0.32759039402008056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,64,power_law_1.01,0.4558784008026123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,64,power_law_1.01,0.10980479717254639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,64,power_law_1.01,0.45204482078552244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,64,power_law_1.01,0.5910719871520996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,64,power_law_1.01,0.12095359563827515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,64,power_law_1.01,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,64,power_law_1.01,0.5717311859130859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,64,power_law_1.01,0.8528256416320801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,64,power_law_1.01,0.09185919761657715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,64,power_law_1.01,0.7816639900207519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,64,power_law_1.01,1.0355199813842773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,64,power_law_1.01,0.09562240242958069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,64,power_law_1.01,1.0330752372741698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,64,power_law_1.01,1.2649727821350099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,64,power_law_1.01,0.10559359788894654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,64,power_law_1.01,1.2756671905517578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,64,power_law_1.01,1.9649856567382813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,64,power_law_1.01,0.10919680595397949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,64,power_law_1.01,2.0031103134155273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,64,power_law_1.01,4.008147048950195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,64,power_law_1.01,0.11429760456085206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,64,power_law_1.01,3.924147033691406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,64,power_law_1.01,0.11068160533905029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,64,power_law_1.01,0.11208319664001465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,64,power_law_1.01,0.11070079803466797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,64,power_law_1.01,0.11598080396652222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,64,power_law_1.01,0.11240960359573364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,64,power_law_1.01,0.1250815987586975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,64,power_law_1.01,0.12375680208206177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,64,power_law_1.01,0.1253119945526123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,64,power_law_1.01,0.13474559783935547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,64,power_law_1.01,0.14503040313720703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,64,power_law_1.01,0.1688256025314331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,64,power_law_1.01,0.19094400405883788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,64,power_law_1.01,0.2465280055999756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,64,power_law_1.01,0.29304959774017336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,64,power_law_1.01,0.3785792112350464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,64,power_law_1.01,0.4683328151702881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,64,power_law_1.01,0.6640511989593506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,64,power_law_1.01,0.8672256469726562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,64,power_law_1.01,1.0870464324951172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,64,power_law_1.01,1.6295232772827148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,64,power_law_1.01,3.177337646484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,64,power_law_1.2,0.1664639949798584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,64,power_law_1.2,0.16838400363922118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,64,power_law_1.2,0.1711680054664612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,64,power_law_1.2,0.09744639992713929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,64,power_law_1.2,0.11818879842758179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,64,power_law_1.2,0.12375680208206177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,64,power_law_1.2,0.14071680307388307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,64,power_law_1.2,0.1563647985458374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,64,power_law_1.2,0.15909119844436645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,64,power_law_1.2,0.16965759992599488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,64,power_law_1.2,0.15461119413375854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,64,power_law_1.2,0.1703552007675171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,64,power_law_1.2,0.15681920051574708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,64,power_law_1.2,0.15261440277099608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,64,power_law_1.2,0.15406719446182252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,64,power_law_1.2,0.17568000555038452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,64,power_law_1.2,0.17624319791793824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,64,power_law_1.2,0.19056639671325684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,64,power_law_1.2,0.19843200445175171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,64,power_law_1.2,0.24459519386291503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,64,power_law_1.2,0.11731840372085571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,64,power_law_1.2,0.2732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,64,power_law_1.2,0.11331199407577515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,64,power_law_1.2,0.35666561126708984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,64,power_law_1.2,0.10674560070037842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,64,power_law_1.2,0.08021759986877441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,64,power_law_1.2,0.40856318473815917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,64,power_law_1.2,0.08586879968643188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,64,power_law_1.2,0.5397247791290283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,64,power_law_1.2,0.09806079864501953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,64,power_law_1.2,0.7085247993469238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,64,power_law_1.2,0.10631680488586426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,64,power_law_1.2,1.1126015663146973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,64,power_law_1.2,0.10756479501724243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,64,power_law_1.2,1.267289638519287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,64,power_law_1.2,0.1272704005241394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,64,power_law_1.2,0.10536320209503174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,64,power_law_1.2,1.5884480476379395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,64,power_law_1.2,0.12487679719924927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,64,power_law_1.2,0.11678719520568848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,64,power_law_1.2,2.477663993835449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,64,power_law_1.2,0.1150272011756897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,64,power_law_1.2,0.11397759914398194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,64,power_law_1.2,0.08388479948043823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,64,power_law_1.2,0.11516159772872925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,64,power_law_1.2,5.7954048156738285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,64,power_law_1.2,0.11603840589523315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,64,power_law_1.2,0.10324480533599853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,64,power_law_1.2,0.11960320472717285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,64,power_law_1.2,0.09032319784164429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,64,power_law_1.2,0.10056320428848267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,64,power_law_1.2,0.12518399953842163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,64,power_law_1.2,0.12629120349884032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,64,power_law_1.2,0.11252479553222657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,64,power_law_1.2,0.12702720165252684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,64,power_law_1.2,0.10601600408554077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,64,power_law_1.2,0.14971519708633424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,64,power_law_1.2,0.1148095965385437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,64,power_law_1.2,0.16040960550308228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,64,power_law_1.2,0.11068160533905029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,64,power_law_1.2,0.11636480093002319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,64,power_law_1.2,0.11044479608535766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,64,power_law_1.2,0.10695040225982666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,64,power_law_1.2,0.12283519506454468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,64,power_law_1.2,0.11961599588394164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,64,power_law_1.2,0.1254271984100342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,64,power_law_1.2,0.13783040046691894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,64,power_law_1.2,0.1476096034049988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,64,power_law_1.2,0.18472319841384888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,64,power_law_1.2,0.20452480316162108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,64,power_law_1.2,0.26883840560913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,64,power_law_1.2,0.3056576013565063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,32,balanced,0.06597333153088887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,64,power_law_1.2,0.4222464084625244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,32,balanced,0.07020266850789388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,32,balanced,0.06959466636180878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,64,power_law_1.2,0.5403264045715332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,32,balanced,0.07584000130494435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,32,balanced,0.0990559955437978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,64,power_law_1.2,0.7703743934631347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,32,balanced,0.12078400452931722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,32,balanced,0.1773759921391805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,64,power_law_1.2,0.9821824073791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,32,balanced,0.16152000427246094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,32,balanced,0.16621333360671997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,64,power_law_1.2,1.217625617980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,32,balanced,0.17036267121632895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,64,power_law_1.2,2.045568084716797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,32,balanced,0.17142399152119955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,32,balanced,0.17113600174585977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,32,balanced,0.16545066237449646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,64,power_law_1.2,4.16429443359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,32,balanced,0.1683893402417501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,32,balanced,0.17165333032608032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,32,balanced,0.17828265825907388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,32,balanced,0.17353065808614096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,32,balanced,0.18162665764490762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,32,balanced,0.18998932838439941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,32,balanced,0.2016693353652954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,32,balanced,0.21461333831151327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,32,balanced,0.2437493403752645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,32,balanced,0.2682080070177714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,32,balanced,0.3267520070075989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,32,balanced,0.4214986562728882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,32,balanced,0.514362653096517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,32,balanced,0.05459199845790863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,32,balanced,0.05491200089454651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,32,balanced,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,32,balanced,0.06314133107662201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,32,balanced,0.06762133538722992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,32,balanced,0.09166933099428813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,32,balanced,0.11443199714024861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,32,balanced,0.11734400192896526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,32,balanced,0.11128532886505127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,32,balanced,0.11611200372378032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,32,balanced,0.6705546379089355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,32,balanced,0.1169546643892924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,32,balanced,0.11392000317573547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,32,balanced,0.11838400363922119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,32,balanced,0.12098667025566101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,32,balanced,0.12777599692344666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,32,balanced,0.12613333264986673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,32,balanced,0.1384266714255015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,32,balanced,0.14313600460688272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,32,balanced,0.1549066702524821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,32,balanced,0.17588265736897787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,32,balanced,0.19716266791025797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,32,balanced,0.8678826491038004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,32,balanced,0.23948800563812256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,32,balanced,0.2813120086987813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,32,balanced,0.36765865484873456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,64,power_law_1.2,0.19341440200805665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,64,power_law_1.2,0.23118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,64,power_law_1.2,0.2825536012649536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,64,power_law_1.2,0.365395188331604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,32,balanced,0.4681440194447835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,32,balanced,0.05630399783452352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,64,power_law_1.2,0.4933055877685547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,32,balanced,0.059338668982187905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,32,balanced,0.058965335289637245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,64,power_law_1.2,0.6470719814300537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,32,balanced,0.06883200009663899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,32,balanced,0.07527466615041097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,64,power_law_1.2,0.876307201385498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,32,balanced,0.10337066650390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,32,balanced,0.12607466181119284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,64,power_law_1.2,1.2547455787658692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,32,balanced,0.11678933103879292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,32,balanced,0.11808000008265178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,64,power_law_1.2,1.493779182434082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,32,balanced,0.11524266997973125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,32,balanced,0.6361226638158163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,32,balanced,1.2876959641774495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,64,power_law_1.2,2.434214401245117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,32,balanced,0.11737599968910217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,32,balanced,0.11573867003122966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,32,balanced,0.1218986709912618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,64,power_law_1.2,4.932096099853515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,32,balanced,0.12282133102416992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,32,balanced,0.12025599678357442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,32,balanced,0.12574932972590128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,32,balanced,0.13317333658536276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,32,balanced,0.13260266184806824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,32,balanced,0.8258453210194906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,32,balanced,0.13902933398882547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,32,balanced,0.1593280037244161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,32,balanced,0.17573867241541544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,32,balanced,0.1993173360824585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,32,balanced,0.23248533407847086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,32,balanced,1.0303093592325847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,32,balanced,0.29478933413823444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,32,balanced,0.396778662999471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,32,balanced,0.5103040138880411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,32,balanced,2.3799360593159995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,32,balanced,0.6651093165079752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,32,balanced,0.8339946269989014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,32,balanced,1.2693813641866047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,32,power_law_1.01,0.12177280187606812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,32,power_law_1.01,0.16379519701004028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,32,power_law_1.01,0.16264959573745727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,32,power_law_1.01,0.08783360123634339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,32,power_law_1.01,0.10354559421539307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,32,power_law_1.01,0.1175104022026062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,32,power_law_1.01,0.12839679718017577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,32,power_law_1.01,0.14209280014038086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,32,power_law_1.01,0.13992960453033448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,32,power_law_1.01,0.15783679485321045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,32,power_law_1.01,0.15287679433822632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,32,power_law_1.01,0.15612159967422484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,32,power_law_1.01,0.16198400259017945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,32,power_law_1.01,0.15889919996261598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,32,power_law_1.01,0.16046719551086425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,32,power_law_1.01,0.17104640007019042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,32,power_law_1.01,0.16538239717483522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,32,power_law_1.01,0.17657599449157715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,32,power_law_1.01,0.1882048010826111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,32,power_law_1.01,0.20861439704895018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,32,power_law_1.01,0.23118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,32,power_law_1.01,0.28296959400177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,32,power_law_1.01,0.33585920333862307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,32,power_law_1.01,0.45416960716247556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,32,balanced,2.4195946057637534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,32,power_law_1.01,0.5519872188568116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,32,power_law_1.01,0.7143487930297852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,32,power_law_1.01,0.9436415672302246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,32,power_law_1.01,1.163590431213379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,32,power_law_1.01,1.7017152786254883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,32,power_law_1.01,3.263488006591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,32,power_law_1.01,0.08974720239639282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,32,power_law_1.01,0.11445759534835816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,32,power_law_1.01,0.11342719793319703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,32,power_law_1.01,0.07245439887046815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,32,power_law_1.01,0.08109440207481385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,32,power_law_1.01,0.09047039747238159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,32,power_law_1.01,0.09246079921722412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,32,power_law_1.01,0.10529919862747192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,32,power_law_1.01,0.10044159889221191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,32,balanced,1.580623944600423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,32,power_law_1.01,0.10957440137863159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,32,power_law_1.01,0.10650880336761474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,32,power_law_1.01,0.1035264015197754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,32,power_law_1.01,0.10620160102844238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,32,power_law_1.01,0.11582720279693604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,32,power_law_1.01,0.11727360486984253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,32,power_law_1.01,0.12165119647979736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,32,power_law_1.01,0.1296895980834961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,32,power_law_1.01,0.14471039772033692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,32,power_law_1.01,0.1539199948310852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,32,power_law_1.01,0.17919360399246215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,32,power_law_1.01,0.20517120361328126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,32,power_law_1.01,0.268723201751709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,32,power_law_1.01,0.32033278942108157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,32,power_law_1.01,0.44391679763793945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,32,power_law_1.01,0.5321792125701904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,32,power_law_1.01,0.7857279777526855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,32,power_law_1.01,0.9799936294555665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,32,power_law_1.01,1.2377344131469727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,32,power_law_1.01,1.875993537902832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,32,power_law_1.01,3.777376174926758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,32,balanced,3.066826820373535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,32,power_law_1.01,0.10432000160217285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,32,power_law_1.01,0.12271360158920289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,32,power_law_1.01,0.11216000318527222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,32,power_law_1.01,0.06910079717636108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,32,power_law_1.2,0.12177280187606812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,32,power_law_1.01,0.0818880021572113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,32,power_law_1.2,0.16775039434432984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,32,power_law_1.01,0.09492480158805847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,32,power_law_1.2,0.14864640235900878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,32,power_law_1.01,0.09682559967041016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,32,power_law_1.2,0.08413439989089966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,32,power_law_1.2,0.1053887963294983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,32,power_law_1.2,0.11406079530715943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,32,power_law_1.2,0.1262719988822937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,32,power_law_1.2,0.13379199504852296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,32,power_law_1.2,0.12773760557174682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,32,power_law_1.2,0.13935999870300292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,32,power_law_1.2,0.1527999997138977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,32,power_law_1.2,0.1396607995033264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,32,power_law_1.2,0.13889919519424437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,32,power_law_1.2,0.14232319593429565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,32,power_law_1.2,0.16189440488815307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,32,power_law_1.2,0.16173440217971802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,32,power_law_1.2,0.17155200242996216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,32,power_law_1.2,0.18896640539169313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,32,power_law_1.2,0.19726719856262206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,32,power_law_1.2,0.2176448106765747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,32,power_law_1.2,0.23578240871429443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,32,power_law_1.2,0.30448639392852783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,32,power_law_1.2,0.3812031984329224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,32,power_law_1.2,0.5136191844940186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,32,power_law_1.2,0.5745215892791748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,32,power_law_1.2,0.8198719978332519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,32,power_law_1.2,1.0238656044006347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,32,power_law_1.2,1.267897605895996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,32,power_law_1.2,2.114431953430176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,32,power_law_1.2,4.38105583190918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,32,power_law_1.2,0.0900223970413208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,32,power_law_1.2,0.11588480472564697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,32,power_law_1.2,0.09193599820137024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,32,power_law_1.2,0.0705407977104187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,32,power_law_1.2,0.0810368001461029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,32,power_law_1.2,0.09076480269432068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,32,power_law_1.2,0.08346880078315735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,32,power_law_1.2,0.10072319507598877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,32,power_law_1.2,0.0938368022441864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,32,power_law_1.2,0.10925439596176148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,32,power_law_1.2,0.10963200330734253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,32,power_law_1.2,0.10922240018844605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,32,power_law_1.2,0.09994239807128906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,32,power_law_1.2,0.10687999725341797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,32,power_law_1.2,0.1198848009109497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,32,power_law_1.2,0.10589439868927002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,32,power_law_1.2,0.10814720392227173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,32,power_law_1.2,0.06832640171051026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,32,power_law_1.2,0.08584960103034973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,32,power_law_1.2,0.08472959995269776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,32,power_law_1.2,0.09353600144386291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,32,power_law_1.2,0.10385279655456543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,32,power_law_1.2,0.10918400287628174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,32,power_law_1.2,0.10877439975738526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,32,power_law_1.2,0.10283520221710205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,32,power_law_1.2,0.11041920185089112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,32,power_law_1.2,0.11090559959411621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,32,power_law_1.2,0.11470719575881957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,32,power_law_1.2,0.11581439971923828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,32,power_law_1.2,0.12210559844970703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,32,power_law_1.2,0.12865279912948607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,32,power_law_1.2,0.1367616057395935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,32,power_law_1.2,0.15378559827804567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,32,power_law_1.2,0.1761791944503784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,32,power_law_1.2,0.20389120578765868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,32,power_law_1.2,0.24666240215301513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,32,power_law_1.2,0.3123071908950806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,32,power_law_1.2,0.40741758346557616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,32,power_law_1.2,0.48970880508422854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,32,power_law_1.2,0.6737088203430176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,32,power_law_1.2,0.8929920196533203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,32,power_law_1.2,1.1682304382324218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,32,power_law_1.2,1.9259647369384765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,32,power_law_1.2,3.7167743682861327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,32,power_law_1.01,0.10510720014572143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,32,power_law_1.01,0.10659199953079224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,32,power_law_1.01,0.11103359460830689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,32,power_law_1.01,0.10673279762268066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,32,power_law_1.01,0.11013120412826538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,32,power_law_1.01,0.1140671968460083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,32,power_law_1.01,0.11679359674453735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,32,power_law_1.01,0.11519360542297363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,32,power_law_1.01,0.11565439701080323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,16,balanced,0.05862933397293091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,32,power_law_1.01,0.12942080497741698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,16,balanced,0.06126399834950765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,16,balanced,0.06329600016276042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,32,power_law_1.01,0.1307263970375061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,16,balanced,0.07841066519419353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,32,power_law_1.01,0.14326399564743042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,16,balanced,0.09470400214195251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,16,balanced,0.1206773320833842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,32,power_law_1.01,0.16367360353469848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,16,balanced,0.16353600223859152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,16,balanced,0.16237866878509521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,32,power_law_1.01,0.19537919759750366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,16,balanced,0.16010666886965433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,32,power_law_1.01,0.24015359878540038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,16,balanced,0.16757865746816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,16,balanced,0.16618133584658304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,32,power_law_1.01,0.2864831924438477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,16,balanced,0.16048533717791238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,32,power_law_1.01,0.3794368028640747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,16,balanced,0.16036267081896463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,16,balanced,0.1630400021870931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,32,power_law_1.01,0.469542407989502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,16,balanced,0.16671466827392578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,32,power_law_1.01,0.6620287895202637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,16,balanced,0.16837332646052042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,32,power_law_1.01,0.819865608215332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,16,balanced,0.17257066567738852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,32,power_law_1.01,0.9942399978637695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,16,balanced,0.18006932735443115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,32,power_law_1.01,1.541977596282959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,16,balanced,0.1885333259900411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,32,power_law_1.01,3.2116607666015624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,16,balanced,0.20517333348592123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,16,balanced,0.22089600563049316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,16,balanced,0.2500266631444295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,16,balanced,0.2791946729024251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,16,balanced,0.34254932403564453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,16,balanced,0.43642131487528485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,32,power_law_1.2,0.11681920289993286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,32,power_law_1.2,0.12409600019454955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,32,power_law_1.2,0.12750719785690307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,16,balanced,0.5540746847788492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,32,power_law_1.2,0.14235520362854004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,32,power_law_1.2,0.15512959957122802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,32,power_law_1.2,0.18947839736938477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,32,power_law_1.2,0.21485440731048583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,32,power_law_1.2,0.2774143934249878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,32,power_law_1.2,0.33935999870300293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,32,power_law_1.2,0.4855679988861084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,16,balanced,0.7130133310953776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,32,power_law_1.2,0.6091968059539795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,32,power_law_1.2,0.8715007781982422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,32,power_law_1.2,1.133568000793457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,32,power_law_1.2,1.3333951950073242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,32,power_law_1.2,2.172102355957031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,32,power_law_1.2,4.497459030151367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,16,balanced,0.050383999943733215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,16,balanced,0.05349866549173991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,16,balanced,0.9143306414286295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,16,balanced,0.054330666859944664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,16,balanced,0.06281599899133046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,16,balanced,0.07170133292675018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,16,balanced,0.09515733520189922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,16,balanced,0.11962133646011353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,16,balanced,0.11571733156840007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,16,balanced,0.11712533235549927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,16,balanced,0.11746666828791301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,16,balanced,0.11746666828791301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,16,balanced,0.11930666367212932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,16,balanced,0.12035199999809265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,16,balanced,0.12265066305796306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,16,balanced,0.12848533193270364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,16,balanced,0.12993599971135458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,16,balanced,0.13569600383440653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,16,balanced,0.1467680037021637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,16,balanced,0.16022933522860208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,16,balanced,0.18477867046991983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,16,balanced,1.3517600695292156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,16,balanced,0.20472000042597452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,16,balanced,0.2528480092684428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,16,balanced,0.2955626646677653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,16,balanced,0.3872213363647461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,16,balanced,0.4999573230743408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,16,balanced,0.053247998158137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,16,balanced,0.05708266794681549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,16,balanced,0.6838613351186117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,16,balanced,0.06107733150323232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,16,balanced,0.06949333349863689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,16,balanced,0.07994666695594788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,16,balanced,0.10466133554776509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,16,balanced,0.12924266854921976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,16,balanced,0.12575466434160867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,16,balanced,0.12642133235931396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,16,balanced,0.12598400314648947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,16,balanced,0.1211946705977122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,16,balanced,0.12563733259836832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,16,balanced,0.12215999762217204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,16,balanced,0.8901546796162924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,16,balanced,0.12794133027394614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,16,balanced,0.13343466321627298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,16,balanced,0.1291253368059794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,16,balanced,2.5515573819478354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,16,balanced,0.1381439963976542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,16,balanced,0.13985066612561545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,16,balanced,0.15055466691652933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,16,balanced,0.168778657913208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,16,balanced,0.185589333375295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,16,balanced,1.0997493267059326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,16,balanced,0.217631995677948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,16,balanced,0.2513440052668254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,16,balanced,0.3222399950027466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,16,balanced,0.43141865730285645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,16,balanced,1.6898719469706218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,16,balanced,0.5671093463897705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,16,power_law_1.01,0.09341440200805665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,16,power_law_1.01,0.12494720220565796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,16,power_law_1.01,0.16853760480880736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,16,power_law_1.01,0.08241919875144958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,16,power_law_1.01,0.09408640265464782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,16,power_law_1.01,0.10885119438171387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,16,power_law_1.01,0.11719679832458496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,16,balanced,0.7354239622751871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,16,power_law_1.01,0.1320896029472351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,16,power_law_1.01,0.14092799425125122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,16,power_law_1.01,0.12255359888076782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,16,power_law_1.01,0.1482432007789612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,16,power_law_1.01,0.13857920169830323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,16,power_law_1.01,0.1472000002861023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,16,power_law_1.01,0.13749760389328003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,16,power_law_1.01,0.15419520139694215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,16,power_law_1.01,0.1504256010055542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,16,power_law_1.01,0.16064640283584594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,16,balanced,0.9167093435923258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,16,power_law_1.01,0.17995519638061525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,16,power_law_1.01,0.17772799730300903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,16,power_law_1.01,0.2077631950378418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,16,balanced,3.283477465311686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,16,power_law_1.01,0.22939519882202147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,16,power_law_1.01,0.2659264087677002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,16,power_law_1.01,0.3239360094070435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,16,power_law_1.01,0.42910079956054686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,16,power_law_1.01,0.5140223979949952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,16,power_law_1.01,0.7216447830200196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,16,power_law_1.01,0.9410112380981446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,16,power_law_1.01,1.1278143882751466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,16,power_law_1.01,1.7806976318359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,16,power_law_1.01,3.325791931152344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,16,balanced,1.4004480044047039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,16,power_law_1.01,0.07447680234909057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,16,power_law_1.01,0.09591680169105529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,16,power_law_1.01,0.11362559795379638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,16,power_law_1.01,0.06869760155677795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,16,power_law_1.01,0.07622399926185608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,16,power_law_1.01,0.08592640161514283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,16,power_law_1.01,0.09299200177192687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,16,power_law_1.01,0.09964159727096558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,16,power_law_1.01,0.09603840112686157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,16,power_law_1.01,0.10112639665603637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,16,power_law_1.01,0.10762879848480225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,16,power_law_1.01,0.10180480480194092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,16,power_law_1.01,0.10842239856719971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,16,power_law_1.01,0.11170560121536255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,16,power_law_1.01,0.1194815993309021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,16,power_law_1.01,0.126636803150177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,16,power_law_1.01,0.13065600395202637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,16,power_law_1.01,0.14266879558563234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,16,power_law_1.01,0.162118399143219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,16,power_law_1.01,0.1921280026435852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,16,power_law_1.01,0.20984320640563964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,16,power_law_1.01,0.2623487949371338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,16,power_law_1.01,0.3251967906951904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,16,power_law_1.01,0.4428351879119873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,16,power_law_1.01,0.5500671863555908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,16,balanced,2.6961545944213867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,16,power_law_1.01,0.7795775890350342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,16,power_law_1.01,1.038041591644287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,16,power_law_1.01,1.2732671737670898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,16,power_law_1.01,1.9907072067260743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,16,power_law_1.01,3.7833152770996095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,16,power_law_1.01,0.08190079927444457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,16,power_law_1.01,0.10267519950866699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,16,power_law_1.01,0.12743680477142333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,16,power_law_1.01,0.0697983980178833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,16,power_law_1.01,0.08039039969444275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,16,power_law_1.01,0.09373440146446228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,16,power_law_1.01,0.1026304006576538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,16,power_law_1.01,0.1085312008857727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,16,power_law_1.01,0.10879360437393189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,16,power_law_1.01,0.11169279813766479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,16,power_law_1.01,0.11019519567489625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,16,power_law_1.01,0.10962560176849365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,16,power_law_1.01,0.1130944013595581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,16,power_law_1.2,0.09456639885902404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,16,power_law_1.01,0.12075519561767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,16,power_law_1.2,0.1268928050994873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,16,power_law_1.01,0.12074240446090698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,16,power_law_1.2,0.10542080402374268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,16,power_law_1.01,0.12674560546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,16,power_law_1.2,0.08418560028076172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,16,power_law_1.01,0.13304320573806763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,16,power_law_1.2,0.09060479998588562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,16,power_law_1.01,0.14062080383300782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,16,power_law_1.2,0.09807360172271729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,16,power_law_1.01,0.15232640504837036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,16,power_law_1.01,0.16986240148544313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,16,power_law_1.2,0.10671360492706299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,16,power_law_1.01,0.20223360061645507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,16,power_law_1.2,0.11830400228500366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,16,power_law_1.01,0.24193921089172363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,16,power_law_1.2,0.12257920503616333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,16,power_law_1.01,0.2970880031585693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,16,power_law_1.2,0.12676479816436767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,16,power_law_1.01,0.39760000705718995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,16,power_law_1.2,0.14442880153656007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,16,power_law_1.2,0.13330559730529784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,16,power_law_1.01,0.482425594329834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,16,power_law_1.2,0.135481595993042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,16,power_law_1.01,0.6923840045928955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,16,power_law_1.2,0.1331007957458496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,16,power_law_1.01,0.8728575706481934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,16,power_law_1.2,0.15824639797210693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,16,power_law_1.01,1.084505558013916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,16,power_law_1.2,0.15775359869003297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,16,power_law_1.01,1.7332799911499024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,16,power_law_1.2,0.16657919883728028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,16,power_law_1.01,3.2423614501953124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,16,power_law_1.2,0.1831488013267517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,16,power_law_1.2,0.1912384033203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,16,power_law_1.2,0.2229248046875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,16,power_law_1.2,0.2315903902053833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,16,power_law_1.2,0.3055295944213867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,16,power_law_1.2,0.36998400688171384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,16,power_law_1.2,0.43991680145263673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,16,power_law_1.2,0.5806464195251465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,16,power_law_1.2,0.8124544143676757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,16,power_law_1.2,1.0337408065795899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,16,power_law_1.2,1.2830656051635743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,16,power_law_1.2,1.9553855895996093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,16,power_law_1.2,0.07429760098457336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,16,power_law_1.2,4.192940902709961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,16,power_law_1.2,0.09596800208091735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,16,power_law_1.2,0.08513919711112976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,16,power_law_1.2,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,16,power_law_1.2,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,16,power_law_1.2,0.08021759986877441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,16,power_law_1.2,0.0837119996547699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,16,power_law_1.2,0.10798720121383668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,16,power_law_1.2,0.0896511971950531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,16,power_law_1.2,0.09314560294151306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,16,power_law_1.2,0.09676160216331482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,16,power_law_1.2,0.07008640170097351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,16,power_law_1.2,0.09696000218391418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,16,power_law_1.2,0.07856640219688416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,16,power_law_1.2,0.09470720291137695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,16,power_law_1.2,0.09749119877815246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,16,power_law_1.2,0.09185919761657715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,16,power_law_1.2,0.09945600032806397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,16,power_law_1.2,0.1087231993675232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,16,power_law_1.2,0.10491520166397095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,16,power_law_1.2,0.10842239856719971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,16,power_law_1.2,0.11043200492858887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,16,power_law_1.2,0.10806399583816528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,16,power_law_1.2,0.11052160263061524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,16,power_law_1.2,0.10714240074157715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,16,power_law_1.2,0.11951359510421752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,16,power_law_1.2,0.10995199680328369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,16,power_law_1.2,0.12725119590759276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,16,power_law_1.2,0.11322239637374878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,16,power_law_1.2,0.13463679552078248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,16,power_law_1.2,0.11427839994430541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,16,power_law_1.2,0.14728319644927979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,16,power_law_1.2,0.12541439533233642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,16,power_law_1.2,0.1641535997390747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,16,power_law_1.2,0.12713600397109986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,16,power_law_1.2,0.18913279771804808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,16,power_law_1.2,0.13277440071105956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,16,power_law_1.2,0.2180799961090088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,16,power_law_1.2,0.14451839923858642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,16,power_law_1.2,0.29247360229492186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,16,power_law_1.2,0.16167039871215821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,16,power_law_1.2,0.34793601036071775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,16,power_law_1.2,0.1824447989463806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,16,power_law_1.2,0.4667967796325684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,16,power_law_1.2,0.2102207899093628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,16,power_law_1.2,0.6049280166625977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,16,power_law_1.2,0.25800960063934325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,16,power_law_1.2,0.882528018951416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,16,power_law_1.2,0.3098112106323242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,16,power_law_1.2,1.1260224342346192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,16,power_law_1.2,0.4324351787567139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,16,power_law_1.2,1.345804786682129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,16,power_law_1.2,0.5297984123229981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,16,power_law_1.2,2.2112831115722655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,16,power_law_1.2,0.7452991962432861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,16,power_law_1.2,4.53309440612793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,16,power_law_1.2,0.9965056419372559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,16,power_law_1.2,1.141324806213379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,16,power_law_1.2,1.9219839096069335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,16,power_law_1.2,3.643743896484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,8,balanced,0.05644266804059347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,8,balanced,0.060991997520128884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,8,balanced,0.06358933448791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,8,balanced,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,8,balanced,0.09690666198730469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,8,balanced,0.12972799936930338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,8,balanced,0.17669866482416788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,8,balanced,0.1647040049235026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,8,balanced,0.16642666856447855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,8,balanced,0.16108266512552896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,8,balanced,0.166101336479187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,8,balanced,0.16488533218701681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,8,balanced,0.16542399923006693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,8,balanced,0.16394666830698648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,8,balanced,0.170522669951121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,8,balanced,0.17031466960906982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,8,balanced,0.17751999696095785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,8,balanced,0.18795732657114664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,8,balanced,0.19750400384267172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,8,balanced,0.2195039987564087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,8,balanced,0.23786665995915732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,8,balanced,0.050426666935284935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,8,balanced,0.055589333176612854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,8,balanced,0.05734399954477946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,8,balanced,0.061530664563179016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,8,balanced,0.07811200122038524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,8,balanced,0.2754506667455037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,8,balanced,0.10215466221173604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,8,balanced,0.12974933783213297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,8,balanced,0.13212266564369202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,8,balanced,0.1306933363278707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,8,balanced,0.31058667103449505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,8,balanced,0.1277653376261393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,8,balanced,0.1302239994208018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,8,balanced,0.13235200444857279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,8,balanced,0.12804266810417175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,8,balanced,0.1348426640033722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,8,balanced,0.14007467031478882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,8,balanced,0.39022934436798096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,8,balanced,0.14338666200637817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,8,balanced,0.15210666259129843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,8,balanced,0.1641866664091746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,8,balanced,0.1755573352177938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,8,balanced,0.5017226537068685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,8,balanced,0.20403200387954712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,8,balanced,0.2278453310330709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,8,balanced,0.27804799874623615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,8,balanced,0.6553386847178141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,8,balanced,0.33155200878779095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,8,balanced,0.4360479911168416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,8,balanced,0.05449066559473673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,8,balanced,0.05922666688760122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,8,balanced,0.06195199986298879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,8,balanced,0.8431573708852133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,8,balanced,0.07195200026035309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,8,balanced,0.08593599994977315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,8,balanced,0.11994666854540507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,8,power_law_1.01,0.08106240034103393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,8,balanced,0.5725760062535604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,8,balanced,0.1508693297704061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,8,power_law_1.01,0.10005120038986207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,8,balanced,0.1476800044377645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,8,balanced,0.14171199997266135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,8,power_law_1.01,0.1282688021659851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,8,balanced,0.14172800381978354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,8,power_law_1.01,0.08103039860725403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,8,balanced,0.1415733297665914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,8,balanced,0.14054933190345764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,8,power_law_1.01,0.09466879963874816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,8,balanced,0.14334400494893393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,8,power_law_1.01,0.10841599702835084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,8,balanced,0.14793599645296732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,8,power_law_1.01,0.11956479549407958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,8,balanced,0.15543466806411743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,8,balanced,0.7840426762898763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,8,balanced,0.15203733245531717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,8,power_law_1.01,0.1247871994972229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,8,balanced,1.045365333557129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,8,balanced,0.1620373328526815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,8,power_law_1.01,0.12669440507888793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,8,balanced,0.16869332393010458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,8,power_law_1.01,0.1301375985145569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,8,power_law_1.01,0.13038079738616942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,8,balanced,0.17774399121602377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,8,power_law_1.01,0.13287680149078368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,8,balanced,0.1987946629524231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,8,power_law_1.01,0.1428607940673828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,8,power_law_1.01,0.14673279523849486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,8,balanced,0.21598933140436807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,8,power_law_1.01,0.15557119846343995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,8,balanced,1.0223573048909504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,8,power_law_1.01,0.15978239774703978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,8,balanced,0.2583306630452474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,8,power_law_1.01,0.16995840072631835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,8,power_law_1.01,0.1819391965866089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,8,balanced,0.29689600070317584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,8,power_law_1.01,0.20064001083374022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,8,power_law_1.01,0.22450559139251708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,8,power_law_1.01,0.24747519493103026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,8,power_law_1.01,0.3006848096847534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,8,balanced,0.3837759892145793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,8,balanced,1.567413330078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,8,power_law_1.01,0.35422720909118655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,8,power_law_1.01,0.45070719718933105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,8,power_law_1.01,0.555398416519165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,8,balanced,1.2624320189158122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,8,power_law_1.01,0.7693120002746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,8,balanced,0.514357328414917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,8,power_law_1.01,1.0019328117370605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,8,power_law_1.01,1.1876992225646972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,8,power_law_1.01,1.9068544387817383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,8,power_law_1.01,3.5157630920410154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,8,balanced,0.6843946774800619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,8,balanced,0.900048017501831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,8,balanced,1.9534187316894531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,8,balanced,1.1061759789784749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,8,balanced,2.9909547170003257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,8,balanced,1.6981706619262695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,8,balanced,3.785130818684896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,8,balanced,3.2794132232666016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,8,power_law_1.01,0.07009279727935791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,8,power_law_1.01,0.0804032027721405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,8,power_law_1.01,0.10384639501571655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,8,power_law_1.01,0.07035520076751708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,8,power_law_1.01,0.07932159900665284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,8,power_law_1.01,0.09348480105400085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,8,power_law_1.01,0.09907839894294738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,8,power_law_1.01,0.1096384048461914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,8,power_law_1.01,0.10868480205535888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,8,power_law_1.01,0.11007360219955445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,8,power_law_1.01,0.10882560014724732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,8,power_law_1.01,0.1233407974243164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,8,power_law_1.01,0.1148095965385437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,8,power_law_1.01,0.12284799814224243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,8,power_law_1.01,0.13860479593276978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,8,power_law_1.01,0.13737599849700927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,8,power_law_1.01,0.14792959690093993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,8,power_law_1.01,0.15768320560455323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,8,power_law_1.01,0.1779584050178528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,8,power_law_1.01,0.2087552070617676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,8,power_law_1.01,0.07141759991645813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,8,power_law_1.01,0.23918719291687013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,8,power_law_1.01,0.08983680009841918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,8,power_law_1.01,0.29901440143585206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,8,power_law_1.01,0.11914880275726318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,8,power_law_1.01,0.3572351932525635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,8,power_law_1.01,0.07221119999885559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,8,power_law_1.01,0.4883584022521973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,8,power_law_1.01,0.08646399974822998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,8,power_law_1.01,0.6593088150024414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,8,power_law_1.01,0.10083199739456176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,8,power_law_1.01,0.8914048194885253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,8,power_law_1.01,0.09955840110778809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,8,power_law_1.01,1.1332672119140625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,8,power_law_1.01,0.11473920345306396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,8,power_law_1.01,1.4064512252807617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,8,power_law_1.01,0.12265599966049194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,8,power_law_1.01,2.1763776779174804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,8,power_law_1.01,0.12289279699325562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,8,power_law_1.01,0.12102400064468384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,8,power_law_1.01,4.378822326660156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,8,power_law_1.01,0.12933759689331054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,8,power_law_1.01,0.12473599910736084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,8,power_law_1.01,0.13097599744796753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,8,power_law_1.01,0.1412992000579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,8,power_law_1.01,0.1459328055381775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,8,power_law_1.01,0.1532415986061096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,8,power_law_1.01,0.16183680295944214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,8,power_law_1.01,0.17550079822540282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,8,power_law_1.01,0.20298879146575927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,8,power_law_1.2,0.0797760009765625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,8,power_law_1.01,0.23364479541778566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,8,power_law_1.01,0.2870271921157837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,8,power_law_1.2,0.09772160053253173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,8,power_law_1.01,0.34875519275665284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,8,power_law_1.2,0.08939520120620728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,8,power_law_1.2,0.08053119778633118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,8,power_law_1.01,0.44745597839355467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,8,power_law_1.2,0.09409919977188111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,8,power_law_1.01,0.5921535968780518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,8,power_law_1.2,0.10261759757995606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,8,power_law_1.01,0.8096575736999512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,8,power_law_1.2,0.1085055947303772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,8,power_law_1.01,1.0722623825073243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,8,power_law_1.2,0.12097920179367065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,8,power_law_1.01,1.2713855743408202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,8,power_law_1.2,0.13674880266189576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,8,power_law_1.01,1.945395278930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,8,power_law_1.2,0.12954239845275878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,8,power_law_1.2,0.13106559514999389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,8,power_law_1.01,3.9892929077148436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,8,power_law_1.2,0.13301119804382325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,8,power_law_1.2,0.1364799976348877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,8,power_law_1.2,0.06679040193557739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,8,power_law_1.2,0.14420479536056519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,8,power_law_1.2,0.08000640273094177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,8,power_law_1.2,0.161190402507782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,8,power_law_1.2,0.08056319952011108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,8,power_law_1.2,0.17288320064544677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,8,power_law_1.2,0.07212799787521362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,8,power_law_1.2,0.16309759616851807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,8,power_law_1.2,0.0770367980003357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,8,power_law_1.2,0.1886016011238098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,8,power_law_1.2,0.08289920091629029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,8,power_law_1.2,0.20241279602050782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,8,power_law_1.2,0.09765759706497193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,8,power_law_1.2,0.22556159496307374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,8,power_law_1.2,0.10364160537719727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,8,power_law_1.2,0.25854079723358153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,8,power_law_1.2,0.10374399423599243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,8,power_law_1.2,0.3054975986480713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,8,power_law_1.2,0.10424319505691529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,8,power_law_1.2,0.3685120105743408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,8,power_law_1.2,0.11381759643554687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,8,power_law_1.2,0.4957888126373291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,8,power_law_1.2,0.11136640310287475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,8,power_law_1.2,0.6191232204437256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,8,power_law_1.2,0.11126400232315063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,8,power_law_1.2,0.8334272384643555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,8,power_law_1.2,0.1206015944480896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,8,power_law_1.2,1.065888023376465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,8,power_law_1.2,0.1293951988220215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,8,power_law_1.2,1.2950207710266113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,8,power_law_1.2,0.13656959533691407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,8,power_law_1.2,1.9948480606079102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,8,power_law_1.2,0.14507520198822021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,8,power_law_1.2,0.16354559659957885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,8,power_law_1.2,3.754022216796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,8,power_law_1.2,0.1803007960319519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,8,power_law_1.2,0.21460480690002443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,8,power_law_1.2,0.24343678951263428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,8,power_law_1.2,0.31137919425964355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,8,power_law_1.2,0.3879040002822876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,8,power_law_1.2,0.5084799766540528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,8,power_law_1.2,0.6608831882476807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,8,power_law_1.2,0.9510848045349121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,8,power_law_1.2,1.2560064315795898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,8,power_law_1.2,1.5952256202697754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,8,power_law_1.2,2.325644874572754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,8,power_law_1.2,5.111148834228516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,8,power_law_1.2,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,8,power_law_1.2,0.08831999897956848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,8,power_law_1.2,0.07826560139656066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,8,power_law_1.2,0.07356799840927124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,8,power_law_1.2,0.08424959778785705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,8,power_law_1.2,0.09350399971008301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,8,power_law_1.2,0.10115840435028076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,8,power_law_1.2,0.11957119703292847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,8,power_law_1.2,0.12345600128173828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,8,power_law_1.2,0.12337919473648071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,8,power_law_1.2,0.12447999715805054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,8,power_law_1.2,0.12803839445114135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,8,power_law_1.2,0.12920960187911987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,8,power_law_1.2,0.1317055940628052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,8,power_law_1.2,0.14172799587249757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,8,power_law_1.2,0.14385279417037963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,8,power_law_1.2,0.1505087971687317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,8,power_law_1.2,0.1724992036819458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,8,power_law_1.2,0.18224639892578126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,4,balanced,0.05898666878541311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,4,balanced,0.06506666541099548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,8,power_law_1.2,0.20780160427093505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,4,balanced,0.06937066713968913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,4,balanced,0.08601066470146179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,8,power_law_1.2,0.2365504026412964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,4,balanced,0.11017066240310669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,4,balanced,0.1425333321094513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,8,power_law_1.2,0.30690560340881345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,4,balanced,0.20402665932973227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,8,power_law_1.2,0.3735487937927246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,4,balanced,0.1925813357035319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,4,balanced,0.1965706745783488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,8,power_law_1.2,0.48824319839477537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,4,balanced,0.19694934288660684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,8,power_law_1.2,0.597049617767334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,4,balanced,0.18742932875951132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,4,balanced,0.19031999508539835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,8,power_law_1.2,0.84202880859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,4,balanced,0.19013333320617676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,8,power_law_1.2,1.130777645111084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,4,balanced,0.19121599197387695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,4,balanced,0.2010293404261271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,8,power_law_1.2,1.4477055549621582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,4,balanced,0.20039467016855875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,8,power_law_1.2,2.126937675476074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,4,balanced,0.20785599946975708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,4,balanced,0.22351467609405518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,8,power_law_1.2,4.283718490600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,4,balanced,0.23522667090098062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,4,balanced,0.2613653341929118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,4,balanced,0.2826240062713623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,4,balanced,0.3388959964116414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,4,balanced,0.05443733433882395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,4,balanced,0.05721066892147064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,4,balanced,0.0609493354956309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,4,balanced,0.06961599985758464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,4,balanced,0.39204267660776776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,4,balanced,0.08983467022577922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,4,balanced,0.11802666385968526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,4,balanced,0.1498186687628428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,4,balanced,0.15009599924087524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,4,balanced,0.1495306690533956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,4,balanced,0.15074666341145834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,4,balanced,0.49794665972391766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,4,balanced,0.15268266201019287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,4,balanced,0.15570666392644247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,4,balanced,0.1564853290716807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,4,balanced,0.16085867087046304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,4,balanced,0.1674826741218567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,4,balanced,0.6562133232752482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,4,balanced,0.17209599415461221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,4,balanced,0.06481066842873891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,4,balanced,0.18200000127156576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,4,balanced,0.06912533442179362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,4,balanced,0.07389866809050243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,4,balanced,0.19731199741363525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,4,balanced,0.0888853371143341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,4,balanced,0.11197866996129353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,4,balanced,0.2116533319155375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,4,balanced,0.16234667102495828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,4,balanced,0.22188800573349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,4,balanced,0.24413333336512247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,4,balanced,0.2059839963912964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,4,balanced,0.20503467321395874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,4,balanced,0.8656960328420004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,4,balanced,0.2002293268839518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,4,balanced,0.2763306697209676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,4,balanced,0.20281066497166952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,4,balanced,0.20002667109171549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,4,balanced,0.2051253318786621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,4,balanced,0.3404426574707031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,4,balanced,0.20901866753896078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,4,balanced,0.2115573287010193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,4,balanced,0.2104746699333191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,4,balanced,0.21975467602411905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,4,balanced,0.40174933274586994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,4,balanced,0.23160000642140707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,4,balanced,1.1224266688028972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,4,balanced,0.24807467063268027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,4,balanced,0.5352693398793539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,4,balanced,0.27433600028355914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,4,balanced,0.2983893354733785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,4,balanced,0.7157173156738281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,4,balanced,0.3577973445256551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,4,balanced,1.3850399653116863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,4,balanced,0.4122026761372884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,4,balanced,0.9857280254364014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,4,balanced,0.537610650062561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,4,balanced,0.7359039783477783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,4,balanced,1.2873013019561768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,4,balanced,2.082085291544596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,4,balanced,0.9643839995066324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,4,balanced,1.594762643178304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,4,balanced,1.2808266480763753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,4,power_law_1.01,0.07347840070724487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,4,power_law_1.01,0.08520960211753845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,4,power_law_1.01,0.11075199842453003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,4,power_law_1.01,0.08398079872131348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,4,power_law_1.01,0.09749119877815246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,4,power_law_1.01,0.11585279703140258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,4,balanced,2.4507840474446616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,4,power_law_1.01,0.1232640027999878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,4,power_law_1.01,0.14001920223236083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,4,power_law_1.01,0.14207359552383422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,4,balanced,1.5779147148132324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,4,power_law_1.01,0.14979840517044068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,4,power_law_1.01,0.15089919567108154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,4,power_law_1.01,0.15623680353164673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,4,power_law_1.01,0.15585919618606567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,4,power_law_1.01,0.16855679750442504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,4,balanced,3.979861259460449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,4,power_law_1.01,0.18640639781951904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,4,power_law_1.01,0.18632320165634156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,4,power_law_1.01,0.19867520332336425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,4,power_law_1.01,0.21161599159240724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,4,power_law_1.01,0.2337023973464966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,4,power_law_1.01,0.2612287998199463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,4,power_law_1.01,0.29791998863220215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,4,power_law_1.01,0.3562943935394287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,4,power_law_1.01,0.4295231819152832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,4,power_law_1.01,0.5537216186523437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,4,power_law_1.01,0.6711359977722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,4,power_law_1.01,0.9462719917297363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,4,power_law_1.01,1.1798784255981445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,4,balanced,2.4192372957865396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,4,power_law_1.01,1.4950143814086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,4,power_law_1.01,2.2688831329345702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,4,power_law_1.01,4.348998260498047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,4,balanced,4.767706553141276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,4,balanced,4.681632041931152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,4,power_law_1.01,0.06476799845695495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,4,power_law_1.01,0.07184640169143677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,4,power_law_1.01,0.08725119829177856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,4,power_law_1.01,0.07349119782447815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,4,power_law_1.01,0.08490880131721497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,4,power_law_1.01,0.09710080027580262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,4,power_law_1.01,0.10598399639129638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,4,power_law_1.01,0.118777596950531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,4,power_law_1.01,0.11826560497283936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,4,power_law_1.01,0.12480000257492066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,4,power_law_1.01,0.12759679555892944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,4,power_law_1.01,0.13084800243377687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,4,power_law_1.01,0.1378175973892212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,4,power_law_1.01,0.1423807978630066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,4,power_law_1.01,0.15948159694671632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,4,power_law_1.01,0.07648640275001525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,4,power_law_1.2,0.07283200025558471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,4,power_law_1.01,0.15466879606246947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,4,power_law_1.01,0.08762879967689514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,4,power_law_1.2,0.08602240085601806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,4,power_law_1.01,0.17534719705581664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,4,power_law_1.01,0.11314560174942016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,4,power_law_1.2,0.07571840286254883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,4,power_law_1.01,0.1927616000175476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,4,power_law_1.01,0.08417919874191285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,4,power_law_1.2,0.08347520232200623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,4,power_law_1.01,0.21100800037384032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,4,power_law_1.01,0.10458879470825196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,4,power_law_1.2,0.0955136001110077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,4,power_law_1.01,0.2506943941116333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,4,power_law_1.01,0.1285248041152954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,4,power_law_1.2,0.11320960521697998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,4,power_law_1.01,0.2812608003616333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,4,power_law_1.01,0.14966399669647218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,4,power_law_1.2,0.12075519561767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,4,power_law_1.01,0.36055679321289064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,4,power_law_1.01,0.1557952046394348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,4,power_law_1.2,0.1401087999343872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,4,power_law_1.01,0.4472640037536621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,4,power_law_1.01,0.1703935980796814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,4,power_law_1.2,0.14086400270462035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,4,power_law_1.01,0.6050367832183838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,4,power_law_1.01,0.16792320013046264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,4,power_law_1.2,0.14304640293121337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,4,power_law_1.01,0.7710720062255859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,4,power_law_1.01,0.16381440162658692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,4,power_law_1.2,0.14736000299453736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,4,power_law_1.01,1.077907180786133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,4,power_law_1.01,0.18365440368652344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,4,power_law_1.2,0.15214719772338867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,4,power_law_1.01,1.3851584434509276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,4,power_law_1.2,0.1564736008644104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,4,power_law_1.01,1.7171648025512696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,4,power_law_1.2,0.16205439567565919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,4,power_law_1.01,2.7305791854858397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,4,power_law_1.2,0.17177599668502808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,4,power_law_1.2,0.17849600315093994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,4,power_law_1.01,5.173318481445312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,4,power_law_1.2,0.20138239860534668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,4,power_law_1.2,0.21799039840698242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,4,power_law_1.2,0.23704960346221923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,4,power_law_1.2,0.2640255928039551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,4,power_law_1.2,0.2990528106689453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,4,power_law_1.2,0.3717695951461792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,4,power_law_1.2,0.4588160037994385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,4,power_law_1.2,0.591315221786499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,4,power_law_1.2,0.7389952182769776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,4,power_law_1.2,1.0032575607299805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,4,power_law_1.2,0.06565759778022766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,4,power_law_1.2,1.2687935829162598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,4,power_law_1.2,0.07343999743461609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,4,power_law_1.2,1.7085184097290038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,4,power_law_1.2,0.06945919990539551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,4,power_law_1.2,2.303225517272949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,4,power_law_1.2,0.07423999905586243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,4,power_law_1.2,4.75140495300293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,4,power_law_1.2,0.0839680016040802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,4,power_law_1.2,0.09545599818229675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,4,power_law_1.2,0.10106240510940552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,4,power_law_1.2,0.11806080341339112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,4,power_law_1.2,0.12031359672546386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,4,power_law_1.2,0.12127360105514526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,4,power_law_1.2,0.12688000202178956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,4,power_law_1.2,0.12899839878082275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,4,power_law_1.2,0.13448959589004517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,4,power_law_1.2,0.1374848008155823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,4,power_law_1.2,0.15216000080108644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,4,power_law_1.2,0.15884799957275392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,4,power_law_1.2,0.17601280212402343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,4,power_law_1.2,0.19108480215072632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,4,power_law_1.2,0.21339519023895265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,4,power_law_1.2,0.2541568040847778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,4,power_law_1.2,0.28689279556274416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,4,power_law_1.2,0.38430080413818357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,4,power_law_1.2,0.4674431800842285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,4,power_law_1.2,0.622982406616211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,4,power_law_1.2,0.7960832118988037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,4,power_law_1.2,1.131270408630371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,4,power_law_1.2,1.4448127746582031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,4,power_law_1.2,1.802150344848633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,4,power_law_1.2,2.8622528076171876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,4,power_law_1.2,5.601004791259766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,4,power_law_1.2,0.07816320061683654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,4,power_law_1.2,0.08519039750099182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,4,power_law_1.2,0.09240319728851318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,4,power_law_1.2,0.08303359746932984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,4,power_law_1.2,0.10341759920120239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,4,power_law_1.2,0.12174079418182374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,4,power_law_1.2,0.12560640573501586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,4,power_law_1.2,0.15593600273132324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,4,power_law_1.2,0.1633471965789795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,4,power_law_1.2,0.1669119954109192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,4,power_law_1.2,0.1678976058959961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,4,power_law_1.2,0.16844160556793214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,4,power_law_1.2,0.16926079988479614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,4,power_law_1.01,0.17809280157089233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,4,power_law_1.2,0.17571200132369996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,4,power_law_1.01,0.17786240577697754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,4,power_law_1.2,0.18881920576095582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,4,power_law_1.01,0.18517760038375855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,4,power_law_1.2,0.1910272002220154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,4,power_law_1.01,0.19173760414123536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,4,power_law_1.2,0.20103681087493896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,4,power_law_1.01,0.20189440250396729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,4,power_law_1.2,0.22412800788879395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,2,balanced,0.0684853345155716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,2,balanced,0.069882666071256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,4,power_law_1.2,0.2439487934112549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,4,power_law_1.01,0.2196671962738037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,2,balanced,0.0755573312441508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,2,balanced,0.09512533744176228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,4,power_law_1.2,0.2834111928939819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,4,power_law_1.01,0.23836159706115723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,2,balanced,0.12583466370900473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,4,power_law_1.01,0.27487359046936033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,4,power_law_1.2,0.3284032106399536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,2,balanced,0.1713386575380961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,2,balanced,0.2329066594441732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,4,power_law_1.01,0.31658239364624025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,4,power_law_1.2,0.40929279327392576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,2,balanced,0.2327679991722107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,4,power_law_1.01,0.39153280258178713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,4,power_law_1.2,0.49176959991455077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,2,balanced,0.23279466231664023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,2,balanced,0.23266132672627768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,4,power_law_1.01,0.46808958053588867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,4,power_law_1.2,0.6676544189453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,2,balanced,0.23479467630386353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,4,power_law_1.01,0.6485824108123779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,4,power_law_1.2,0.8576128005981445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,2,balanced,0.2397973338762919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,4,power_law_1.01,0.7747456073760987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,4,power_law_1.2,1.2207615852355957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,2,balanced,0.2406933307647705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,2,balanced,0.24197334051132202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,4,power_law_1.01,1.0666239738464356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,4,power_law_1.2,1.5052160263061523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,2,balanced,0.2537546753883362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,4,power_law_1.01,1.4629055976867675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,4,power_law_1.2,1.7648704528808594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,2,balanced,0.25491732358932495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,4,power_law_1.01,1.6852672576904297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,4,power_law_1.2,2.834284782409668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,2,balanced,0.2674986720085144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,4,power_law_1.01,2.676460838317871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,2,balanced,0.2893226742744446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,4,power_law_1.2,5.413241577148438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,2,balanced,0.30696000655492145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,4,power_law_1.01,5.252799987792969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,2,balanced,0.3496853510538737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,2,balanced,0.38570133845011395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,2,balanced,0.46697600682576496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,2,balanced,0.5390880107879639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,2,balanced,0.7001012961069742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,2,balanced,0.0625546673933665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,2,balanced,0.06557333469390869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,2,balanced,0.07162666817506154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,2,balanced,0.08829866846402486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,2,balanced,0.1118986705938975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,2,balanced,0.16077867150306702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,2,balanced,0.21222400665283203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,2,balanced,0.9464373588562012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,2,balanced,0.20916267236073813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,2,balanced,0.21593600511550903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,2,balanced,0.21919999519983926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,2,balanced,0.21461333831151327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,2,balanced,0.22087466716766357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,2,balanced,0.2222879926363627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,2,balanced,0.22504534324010214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,2,balanced,0.23625600337982178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,2,balanced,1.2668533325195312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,2,balanced,0.23669866720835367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,2,balanced,0.25066133340199787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,2,balanced,0.27400533358256024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,2,balanced,0.29291733105977374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,2,balanced,0.3418346643447876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,2,balanced,0.38178133964538574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,2,balanced,1.6534934043884277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,2,balanced,0.4757653474807739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,2,balanced,0.5618240038553873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,2,balanced,0.7492480278015137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,2,balanced,2.0632960001627603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,2,balanced,1.034010648727417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,2,balanced,1.4087093671162922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,2,balanced,3.1390345891316733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,2,balanced,1.8692480723063152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,2,power_law_1.01,0.0717311978340149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,2,power_law_1.01,0.07847679853439331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,2,power_law_1.01,0.09207040071487427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,2,power_law_1.01,0.0664896011352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,2,power_law_1.01,0.09422079920768738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,2,power_law_1.01,0.07217280268669128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,2,power_law_1.01,0.11459840536117553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,2,power_law_1.01,0.08581759929656982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,2,balanced,2.3253226280212402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,2,power_law_1.01,0.13118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,2,power_law_1.01,0.08498560190200806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,2,power_law_1.01,0.1403839945793152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,2,power_law_1.01,0.10841599702835084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,2,power_law_1.01,0.16508159637451172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,2,power_law_1.01,0.12128000259399414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,2,power_law_1.01,0.13248640298843384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,2,power_law_1.01,0.17277439832687377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,2,power_law_1.01,0.15676159858703614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,2,power_law_1.01,0.17487360239028932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,2,power_law_1.01,0.16417920589447021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,2,power_law_1.01,0.1835904002189636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,2,power_law_1.01,0.1694975972175598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,2,power_law_1.01,0.1923200011253357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,2,power_law_1.01,0.1732159972190857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,2,power_law_1.01,0.19263360500335694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,2,power_law_1.01,0.18040319681167602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,2,power_law_1.01,0.20094718933105468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,2,power_law_1.01,0.1923840045928955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,2,power_law_1.01,0.21715199947357178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,2,power_law_1.01,0.19614720344543457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,2,power_law_1.01,0.2202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,2,power_law_1.01,0.20612480640411376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,2,power_law_1.01,0.2422976016998291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,2,power_law_1.01,0.2127295970916748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,2,power_law_1.01,0.262009596824646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,2,power_law_1.01,0.23048319816589355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,2,power_law_1.01,0.287225604057312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,2,balanced,3.5835466384887695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,2,power_law_1.01,0.2543168067932129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,2,power_law_1.01,0.34639999866485593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,2,power_law_1.01,0.28095359802246095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,2,power_law_1.01,0.3849855899810791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,2,power_law_1.01,0.33315200805664064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,2,power_law_1.01,0.48817920684814453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,2,power_law_1.01,0.3869568109512329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,2,power_law_1.01,0.5675903797149658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,2,balanced,6.0758717854817705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,2,power_law_1.01,0.49399681091308595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,2,power_law_1.01,0.7788479804992676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,2,power_law_1.01,0.6149824142456055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,2,power_law_1.01,0.9631296157836914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,2,power_law_1.01,0.8486144065856933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,2,power_law_1.01,1.3173888206481934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,2,power_law_1.01,1.0749695777893067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,2,power_law_1.01,1.723302459716797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,2,power_law_1.01,1.500870418548584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,2,power_law_1.01,2.1063936233520506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,2,power_law_1.01,1.910905647277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,2,power_law_1.01,3.2158145904541016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,2,power_law_1.01,2.3546112060546873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,2,power_law_1.01,6.22935676574707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,2,power_law_1.01,3.7727359771728515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,2,power_law_1.01,7.324851226806641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,2,balanced,6.996005376180013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,8,balanced,0.03997333347797394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,8,balanced,0.04104000081618627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,8,balanced,0.03959999978542328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,8,balanced,0.040922666589419045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,8,balanced,0.04252799848715464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,8,balanced,0.044031997521718345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,8,balanced,0.04288533329963684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,8,balanced,0.043023998538653054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,8,balanced,0.04246933261553446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,8,balanced,0.04275199770927429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,8,balanced,0.042992000778516136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,8,balanced,0.042730664213498436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,8,balanced,0.04327466587225596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,8,balanced,0.04298666616280874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,8,balanced,0.046367997924486794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,8,balanced,0.04658666749795278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,8,balanced,0.046538665890693665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,8,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,8,balanced,0.04894400139649709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,8,balanced,0.053504000107447304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,8,balanced,0.05685866872469584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,8,balanced,0.06543999910354614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,8,balanced,0.07361599802970886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,2,power_law_1.2,0.07131519913673401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,8,balanced,0.09956266482671101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,2,power_law_1.2,0.07733759880065919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,8,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,8,balanced,0.03698666642109553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,2,power_law_1.2,0.08173440098762512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,8,balanced,0.03700799991687139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,8,balanced,0.11493333180745442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,8,balanced,0.03860799968242645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,2,power_law_1.2,0.0939903974533081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,8,balanced,0.03898133337497711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,2,power_law_1.2,0.10714240074157715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,8,balanced,0.03905600061019262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,8,balanced,0.150325338045756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,8,balanced,0.0408746674656868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,2,power_law_1.2,0.1269376039505005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,8,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,2,power_law_1.2,0.13521920442581176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,8,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,2,power_law_1.2,0.06581119894981384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,8,balanced,0.040149333576361336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,2,power_law_1.2,0.16592639684677124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,8,balanced,0.17710934082667032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,2,power_law_1.2,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,8,balanced,0.04152533411979675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,8,balanced,0.041189332803090416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,2,power_law_1.2,0.16513919830322266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,2,power_law_1.2,0.07418879866600037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,8,balanced,0.04237333436806997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,2,power_law_1.2,0.16841599941253663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,2,power_law_1.2,0.0806335985660553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,8,balanced,0.04142399877309799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,8,balanced,0.21446933348973593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,8,balanced,0.04530133306980133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,2,power_law_1.2,0.17681280374526978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,2,power_law_1.2,0.10707199573516846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,8,balanced,0.04488533238569895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,2,power_law_1.2,0.18766720294952394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,2,power_law_1.2,0.11782399415969849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,8,balanced,0.048986668388048805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,2,power_law_1.2,0.19026559591293335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,2,power_law_1.2,0.12789759635925294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,8,balanced,0.048800001541773476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,8,balanced,0.05298133194446564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,2,power_law_1.2,0.19793280363082885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,2,power_law_1.2,0.15620479583740235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,8,balanced,0.3162933389345805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,8,balanced,0.0572320024172465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,2,power_law_1.2,0.2094655990600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,2,power_law_1.2,0.16329599618911744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,8,balanced,0.06338133414586385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,2,power_law_1.2,0.2277695894241333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,2,power_law_1.2,0.16314879655838013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,8,balanced,0.0786293347676595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,2,power_law_1.2,0.2524928092956543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,2,power_law_1.2,0.16651519536972045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,2,power_law_1.2,0.2662015914916992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,2,power_law_1.2,0.1753216028213501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,8,balanced,0.10378666718800862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,2,power_law_1.2,0.298905611038208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,2,power_law_1.2,0.18248319625854492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,8,balanced,0.1423679987589518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,2,power_law_1.2,0.3524224042892456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,2,power_law_1.2,0.18812799453735352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,2,power_law_1.2,0.40033278465270994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,2,power_law_1.2,0.20666239261627198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,8,balanced,0.1728853384653727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,2,power_law_1.2,0.20879359245300294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,2,power_law_1.2,0.4948607921600342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,8,balanced,0.5904693206151327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,2,power_law_1.2,0.2302783966064453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,2,power_law_1.2,0.5922944068908691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,2,power_law_1.2,0.2578432083129883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,2,power_law_1.2,0.7775231838226319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,8,balanced,0.23953600724538168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,2,power_law_1.2,0.2900991916656494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,2,power_law_1.2,0.9806912422180176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,2,power_law_1.2,0.3428992033004761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,2,power_law_1.2,1.401465606689453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,2,power_law_1.2,0.40923519134521485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,2,power_law_1.2,1.8677888870239259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,8,balanced,0.29732267061869305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,2,power_law_1.2,0.5175104141235352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,2,power_law_1.2,2.090176010131836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,2,power_law_1.2,0.6403903961181641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,2,power_law_1.2,3.242892837524414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,2,power_law_1.2,0.8502592086791992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,8,balanced,0.3628480037053426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,2,power_law_1.2,6.733241271972656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,2,power_law_1.2,1.0685952186584473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,2,power_law_1.2,1.502732753753662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,2,power_law_1.2,1.9908735275268554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,8,balanced,0.548853317896525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,2,power_law_1.2,2.3297279357910154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,2,power_law_1.2,3.6468990325927733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,2,power_law_1.2,7.37391357421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,8,balanced,1.0586880048116047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,8,balanced,0.03612266729275385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,8,balanced,0.03692800054947535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,8,balanced,0.036831999818483986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,8,balanced,0.03899733225504557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,8,balanced,0.038575999438762665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,8,balanced,0.038975998759269714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,8,balanced,0.04092800120512644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,8,balanced,0.04081599911053976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,8,balanced,0.039061332742373146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,8,balanced,0.04061333338419596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,8,balanced,0.040522667268911995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,8,balanced,0.041221333046754204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,8,balanced,0.040005333721637726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,8,balanced,0.04065066576004028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,8,balanced,0.04489600161711375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,8,balanced,0.019306667149066925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,8,balanced,0.04453866680463155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,8,balanced,0.020661332954963047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,8,balanced,0.04496000210444132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,8,balanced,0.02067199970285098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,8,balanced,0.0489333321650823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,8,balanced,0.020453333854675293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,8,balanced,0.02025066688656807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,8,balanced,0.05235200126965841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,8,balanced,0.022704000274340313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,8,balanced,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,8,balanced,0.04062400013208389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,8,balanced,0.060773332913716636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,8,balanced,0.038986665507157646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,8,power_law_1.01,0.039315199851989745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,8,balanced,0.03179733455181122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,8,power_law_1.01,0.039647999405860904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,8,balanced,0.07349333167076111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,8,balanced,0.031632001201311745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,8,power_law_1.01,0.039417600631713866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,8,balanced,0.032698666055997215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,8,power_law_1.01,0.04136320054531097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,8,balanced,0.08360000451405843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,8,balanced,0.03319466610749563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,8,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,8,balanced,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,8,power_law_1.01,0.04217599928379059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,8,balanced,0.10955199599266052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,8,balanced,0.032842665910720825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,8,power_law_1.01,0.041305598616600034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,8,balanced,0.03631466627120972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,8,power_law_1.01,0.04184960126876831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,8,power_law_1.01,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,8,balanced,0.034474665919939675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,8,power_law_1.01,0.04168959856033325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,8,power_law_1.01,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,8,balanced,0.1269706686337789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,8,power_law_1.01,0.04162560105323791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,8,balanced,0.03619733452796936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,8,power_law_1.01,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,8,power_law_1.01,0.041843199729919435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,8,power_law_1.01,0.036831998825073244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,8,balanced,0.04079466561476389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,8,power_law_1.01,0.04254080057144165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,8,power_law_1.01,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,8,balanced,0.0430026650428772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,8,balanced,0.17180800437927246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,8,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,8,power_law_1.01,0.04254719913005829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,8,balanced,0.05517866710821787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,8,power_law_1.01,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,8,power_law_1.01,0.04262399971485138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,8,power_law_1.01,0.038624000549316403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,8,balanced,0.060453335444132485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,8,power_law_1.01,0.03888640105724335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,8,balanced,0.07970666885375977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,8,power_law_1.01,0.038950398564338684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,8,balanced,0.2133973240852356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,8,power_law_1.01,0.03940480053424835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,8,balanced,0.08957866827646892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,8,power_law_1.01,0.039417600631713866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,8,power_law_1.01,0.040531200170516965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,8,power_law_1.01,0.040627199411392215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,8,balanced,0.12446932991345723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,8,power_law_1.01,0.04629760086536407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,8,balanced,0.25549866755803424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,8,power_law_1.01,0.04739840030670166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,8,power_law_1.01,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,8,balanced,0.15034666657447815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,8,power_law_1.01,0.05335680246353149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,8,power_law_1.01,0.060889601707458496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,8,power_law_1.01,0.0712768018245697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,8,power_law_1.01,0.08149759769439698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,8,balanced,0.21218132972717285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,8,power_law_1.01,0.11808639764785767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,8,power_law_1.01,0.1480512022972107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,8,balanced,0.3805493513743083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,8,power_law_1.01,0.20543360710144043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,8,power_law_1.01,0.26988160610198975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,8,balanced,0.27217066287994385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,8,power_law_1.01,0.3652415990829468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,8,power_law_1.01,0.5051648139953613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,8,power_law_1.01,0.6117184162139893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,8,power_law_1.01,0.9342592239379883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,8,power_law_1.01,1.8938751220703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,8,balanced,0.3337920109430949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,8,balanced,0.5187893311182658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,8,balanced,0.717519998550415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,8,balanced,1.004853328069051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,8,power_law_1.01,0.03544319868087768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,8,power_law_1.01,0.03504000008106232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,8,power_law_1.01,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,8,power_law_1.01,0.03681280016899109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,8,power_law_1.01,0.037459200620651244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,8,power_law_1.01,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,8,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,8,power_law_1.01,0.03779839873313904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,8,power_law_1.01,0.017715199291706084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,8,power_law_1.01,0.03813120126724243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,8,power_law_1.01,0.019526399672031403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,8,power_law_1.01,0.03834879994392395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,8,power_law_1.01,0.018400000035762788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,8,power_law_1.01,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,8,power_law_1.01,0.018508799374103546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,8,power_law_1.01,0.040889599919319154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,8,power_law_1.01,0.019500799477100372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,8,power_law_1.01,0.04066559970378876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,8,power_law_1.01,0.020960000157356263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,8,power_law_1.01,0.04066559970378876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,8,power_law_1.01,0.038278400897979736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,8,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,8,power_law_1.01,0.03758080005645752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,8,power_law_1.01,0.04445439875125885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,8,power_law_1.01,0.04852479994297028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,8,power_law_1.01,0.02913280129432678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,8,power_law_1.01,0.055180799961090085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,8,power_law_1.01,0.02963840067386627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.01,0.060102397203445436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,8,power_law_1.01,0.03030399978160858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.01,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,8,power_law_1.01,0.030668801069259642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.01,0.08439040184020996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,8,power_law_1.01,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.01,0.10380799770355224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,8,power_law_1.01,0.0313728004693985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.01,0.1266495943069458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,8,power_law_1.01,0.03335680067539215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,8,power_law_1.01,0.03296639919281006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.01,0.16432640552520753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,8,power_law_1.01,0.03395200073719025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.01,0.21543679237365723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,8,power_law_1.01,0.037785598635673524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.01,0.3178688049316406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.01,0.042099198698997496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.01,0.3895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,8,power_law_1.2,0.039961600303649904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.01,0.053939199447631835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,8,power_law_1.01,0.5025279998779297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,8,power_law_1.2,0.03991680145263672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.01,0.060115200281143186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,8,power_law_1.01,0.728275203704834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,8,power_law_1.2,0.0395904004573822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.01,0.07963520288467407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,8,power_law_1.2,0.040582400560379026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,8,power_law_1.01,1.4942208290100099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.01,0.08917120099067688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,8,power_law_1.2,0.04132480025291443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,8,power_law_1.01,0.046387198567390445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.01,0.12487679719924927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,8,power_law_1.2,0.04178560078144074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,8,power_law_1.01,0.048665601015090945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.01,0.1509951949119568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,8,power_law_1.2,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,8,power_law_1.01,0.05050240159034729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,8,power_law_1.2,0.04092159867286682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.01,0.2118272066116333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,8,power_law_1.01,0.05623679757118225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,8,power_law_1.2,0.04154880046844482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.01,0.27320959568023684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,8,power_law_1.01,0.06263039708137512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,8,power_law_1.2,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,8,power_law_1.01,0.3352895975112915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,8,power_law_1.01,0.07861760258674622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,8,power_law_1.2,0.04120959937572479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,8,power_law_1.01,0.5211967945098877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,8,power_law_1.01,0.08748160004615783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,8,power_law_1.2,0.04284160137176514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,8,power_law_1.01,1.0095359802246093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,8,power_law_1.01,0.115283203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,8,power_law_1.2,0.043833601474761966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,8,power_law_1.01,0.135481595993042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,8,power_law_1.2,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,8,power_law_1.01,0.17456640005111695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,8,power_law_1.2,0.048390400409698484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,8,power_law_1.2,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,8,power_law_1.01,0.22115840911865234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,8,power_law_1.2,0.04987519979476929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,8,power_law_1.01,0.33447680473327634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,8,power_law_1.2,0.05663359761238098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,8,power_law_1.01,0.4078335762023926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,8,power_law_1.2,0.06297600269317627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,8,power_law_1.01,0.4972799777984619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,8,power_law_1.2,0.08009600043296813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,8,power_law_1.01,0.8431872367858887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,8,power_law_1.2,0.08822399973869324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,8,power_law_1.01,1.641881561279297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,8,power_law_1.2,0.11141760349273681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,8,power_law_1.2,0.13962240219116212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,8,power_law_1.2,0.18729599714279174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,8,power_law_1.2,0.22458240985870362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,8,power_law_1.2,0.3223104000091553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,8,power_law_1.2,0.4216639995574951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,8,power_law_1.2,0.54301438331604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,8,power_law_1.2,0.8168383598327636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,8,power_law_1.2,1.6558975219726562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,8,power_law_1.2,0.03511680066585541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,8,power_law_1.2,0.035046398639678955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,8,power_law_1.2,0.035123199224472046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,8,power_law_1.2,0.036294400691986084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,8,power_law_1.2,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,8,power_law_1.2,0.040320000052452086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,8,power_law_1.2,0.038022398948669434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,8,power_law_1.2,0.03409920036792755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,8,power_law_1.2,0.038624000549316403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,8,power_law_1.2,0.03573760092258453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,8,power_law_1.2,0.03842560052871704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,8,power_law_1.2,0.03478400111198425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,8,power_law_1.2,0.038771200180053714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,8,power_law_1.2,0.03923200070858002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,8,power_law_1.2,0.03758719861507416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,8,power_law_1.2,0.040524798631668094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,8,power_law_1.2,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,8,power_law_1.2,0.0404992014169693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,8,power_law_1.2,0.03758080005645752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,8,power_law_1.2,0.040320000052452086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,8,power_law_1.2,0.03775359988212586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,8,power_law_1.2,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,8,power_law_1.2,0.03924480080604553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,8,power_law_1.2,0.018777599930763243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,8,power_law_1.2,0.04755200147628784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,8,power_law_1.2,0.03856639862060547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,8,power_law_1.2,0.019782400131225585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,8,power_law_1.2,0.04933120012283325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,8,power_law_1.2,0.03918719887733459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,8,power_law_1.2,0.019193600118160247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,8,power_law_1.2,0.056806397438049314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,8,power_law_1.2,0.038489601016044615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,8,power_law_1.2,0.018956799805164338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,8,power_law_1.2,0.06154879927635193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,8,power_law_1.2,0.040428799390792844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,8,power_law_1.2,0.018515199422836304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,8,power_law_1.2,0.07641599774360656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,8,power_law_1.2,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,8,power_law_1.2,0.021414400637149812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,8,power_law_1.2,0.08917760252952575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,8,power_law_1.2,0.0416703999042511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,8,power_law_1.2,0.038387200236320494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,4,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,8,power_law_1.2,0.12922879457473754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,8,power_law_1.2,0.04620159864425659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,8,power_law_1.2,0.03848319947719574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,4,balanced,0.039061332742373146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,8,power_law_1.2,0.15415680408477783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,8,power_law_1.2,0.047737601399421695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,8,power_law_1.2,0.030694401264190672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,4,balanced,0.04197333256403605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,8,power_law_1.2,0.21639039516448974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,8,power_law_1.2,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,8,power_law_1.2,0.050374400615692136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,4,balanced,0.04341333111127218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,8,power_law_1.2,0.25719680786132815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,8,power_law_1.2,0.030457600951194763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,4,balanced,0.04114133367935816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,8,power_law_1.2,0.05790079832077026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,4,balanced,0.04326933125654856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,8,power_law_1.2,0.38606719970703124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,8,power_law_1.2,0.031097599864006044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.2,0.06307200193405152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,4,balanced,0.04279999931653341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,8,power_law_1.2,0.032313600182533264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,8,power_law_1.2,0.509222412109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.2,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,4,balanced,0.04108799993991852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,8,power_law_1.2,0.03128319978713989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,8,power_law_1.2,0.6257279872894287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.2,0.08382080197334289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,4,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,8,power_law_1.2,0.0342848002910614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.2,0.10807679891586304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,4,balanced,0.04371733466784159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,8,power_law_1.2,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.2,0.12481919527053834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,4,balanced,0.0432533323764801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,8,power_law_1.2,0.03438720107078552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.2,0.17864960432052612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,4,balanced,0.04250133534272512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,8,power_law_1.2,0.03882879912853241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.2,0.2212224006652832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,4,balanced,0.0424586683511734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.2,0.042822399735450746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.2,0.2980031967163086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,4,balanced,0.04304533203442892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.2,0.05441920161247253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.2,0.40705280303955077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,4,balanced,0.046767999728520714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.2,0.060729598999023436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,4,balanced,0.04587199787298838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,8,power_law_1.2,0.5065792083740235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.2,0.07765759825706482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,4,balanced,0.0489279975493749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,8,power_law_1.2,0.7799615859985352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.2,0.08888959884643555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,4,balanced,0.051728000243504844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.2,0.12382719516754151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,8,power_law_1.2,1.451315212249756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.2,0.15240960121154784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,4,balanced,0.052655999859174095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.2,0.2130176067352295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,4,balanced,0.057402665416399636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.2,0.2727936029434204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,4,balanced,0.06172266602516174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,8,power_law_1.2,0.33631999492645265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,8,power_law_1.2,0.5200384140014649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,4,balanced,0.07282133400440216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,8,power_law_1.2,1.0047295570373536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,4,balanced,0.08356266220410664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,4,balanced,0.1107306679089864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,4,balanced,0.12622933586438498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,4,balanced,0.17545600732167563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,4,balanced,0.21126399437586466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,4,balanced,0.2571679949760437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,4,balanced,0.3800266583760579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,4,balanced,0.716474692026774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,4,balanced,0.0340639998515447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,4,balanced,0.03497066597143809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,4,balanced,0.03691199918588003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,4,balanced,0.03892799963553747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,4,balanced,0.03899200012286504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,4,balanced,0.039120001097520195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,4,balanced,0.042693331837654114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,4,balanced,0.03912533322970072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,4,balanced,0.04080000023047129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,4,balanced,0.04088533421357473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,4,balanced,0.0406986673672994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,4,balanced,0.04248000184694926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,4,power_law_1.01,0.039238399267196654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,4,balanced,0.040991999208927155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,4,power_law_1.01,0.03927040100097656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,4,balanced,0.04111466556787491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,4,power_law_1.01,0.04069119989871979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,4,balanced,0.04671466847260793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,4,power_law_1.01,0.042387199401855466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,4,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,4,balanced,0.046037331223487854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,4,power_law_1.01,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,4,balanced,0.04711999992529551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,4,power_law_1.01,0.041740798950195314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,4,balanced,0.050810664892196655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,4,power_law_1.01,0.04225279986858368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,4,balanced,0.0569653312365214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,4,power_law_1.01,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,4,balanced,0.06163200239340464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,4,power_law_1.01,0.04280959963798523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,4,power_law_1.01,0.04227840006351471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,4,balanced,0.07089599967002869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,4,power_law_1.01,0.04408960044384003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,4,balanced,0.09238933523495992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,4,power_law_1.01,0.042796799540519716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,4,power_law_1.01,0.04385280013084412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,4,balanced,0.11788800358772278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,4,power_law_1.01,0.047091200947761536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,4,power_law_1.01,0.048819199204444885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,4,power_law_1.01,0.05023360252380371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,4,balanced,0.15821333726247153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,4,power_law_1.01,0.05807999968528747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,4,power_law_1.01,0.060787200927734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,4,power_law_1.01,0.07376639842987061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,4,balanced,0.19247466325759888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,4,power_law_1.01,0.08058239817619324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,4,power_law_1.01,0.10160640478134156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,4,power_law_1.01,0.12735359668731688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,4,power_law_1.01,0.20049920082092285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,4,balanced,0.27020267645517987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,4,power_law_1.01,0.21040000915527343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,4,power_law_1.01,0.3221247911453247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,4,power_law_1.01,0.41989760398864745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,4,power_law_1.01,0.4881152153015137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,4,balanced,0.33905601501464844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,4,power_law_1.01,0.7887104034423829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,4,power_law_1.01,1.3293696403503419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,8,power_law_1.2,0.9748096466064453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,4,balanced,0.4145919879277547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,8,power_law_1.2,1.9001407623291016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,4,balanced,0.6380106608072916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,4,power_law_1.01,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,4,power_law_1.01,0.035795199871063235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,4,power_law_1.01,0.03562879860401154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,4,power_law_1.01,0.03739520013332367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,4,power_law_1.01,0.03877759873867035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,4,power_law_1.01,0.03841919898986816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,4,balanced,1.2397706508636475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,4,power_law_1.01,0.04020479917526245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,4,power_law_1.01,0.03903999924659729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,4,power_law_1.01,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,4,power_law_1.01,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,4,power_law_1.01,0.04032639861106872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,4,power_law_1.01,0.04142079949378967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,4,power_law_1.01,0.041228801012039185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,4,power_law_1.01,0.04286080002784729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,4,power_law_1.01,0.04577920138835907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,4,power_law_1.01,0.050483202934265135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,4,power_law_1.01,0.052134400606155394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,4,power_law_1.01,0.058195197582244874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,4,power_law_1.01,0.06476160287857055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,4,power_law_1.01,0.07523199915885925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,4,power_law_1.01,0.09715200066566468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,4,power_law_1.01,0.1283519983291626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,4,power_law_1.01,0.1634559988975525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,4,power_law_1.01,0.23236479759216308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,4,power_law_1.01,0.26841599941253663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,4,power_law_1.01,0.39429121017456054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,4,power_law_1.01,0.5256896018981934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,4,power_law_1.01,0.6454207897186279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,4,power_law_1.01,1.0049728393554687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,4,power_law_1.01,2.3322111129760743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,4,power_law_1.2,0.038022398948669434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,4,power_law_1.2,0.038815999031066896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,4,power_law_1.2,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,4,power_law_1.2,0.043315199017524716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,4,power_law_1.2,0.04215039908885956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,4,power_law_1.2,0.04108160138130188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,4,power_law_1.2,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,4,power_law_1.2,0.04215039908885956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,4,power_law_1.2,0.036908799409866334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,4,power_law_1.2,0.042803201079368594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,4,power_law_1.2,0.035596799850463864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,4,power_law_1.2,0.041740798950195314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,4,power_law_1.2,0.04182400107383728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,4,power_law_1.2,0.038073599338531494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,4,power_law_1.2,0.041433599591255185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,4,power_law_1.2,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,4,power_law_1.2,0.04297600090503693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,4,power_law_1.2,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,4,power_law_1.2,0.043270400166511534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,4,power_law_1.2,0.03891839981079102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,4,power_law_1.2,0.04366079866886139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,4,power_law_1.2,0.039680001139640805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,4,power_law_1.2,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,4,power_law_1.2,0.041459199786186215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,4,power_law_1.2,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,4,power_law_1.2,0.04009599983692169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,4,power_law_1.2,0.05118079781532288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,4,power_law_1.2,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,4,power_law_1.2,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,4,power_law_1.2,0.04111360013484955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,4,power_law_1.2,0.06270080208778381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,4,power_law_1.2,0.041145598888397215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,4,power_law_1.2,0.07669119834899903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,4,power_law_1.2,0.042444801330566405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,8,balanced,0.048528000712394714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,4,power_law_1.2,0.08732159733772278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,4,power_law_1.2,0.04707199931144714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,8,balanced,0.049653331438700356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,4,power_law_1.2,0.12098560333251954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,4,power_law_1.2,0.048511999845504764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,8,balanced,0.050293331344922386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,4,power_law_1.2,0.050271999835968015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,4,power_law_1.2,0.1424064040184021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,8,balanced,0.05089599887530009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,4,power_law_1.2,0.058182400465011594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,4,power_law_1.2,0.17747199535369873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,8,balanced,0.05130666494369507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,8,balanced,0.05035733183224996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,4,power_law_1.2,0.06401280164718628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,4,power_law_1.2,0.24927999973297119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,8,balanced,0.0510453333457311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,4,power_law_1.2,0.08266239762306213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,4,power_law_1.2,0.3612735986709595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,8,balanced,0.05136000116666158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,4,power_law_1.2,0.11183359622955322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,4,power_law_1.2,0.39782400131225587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,8,balanced,0.05041599770387014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,4,power_law_1.2,0.13969919681549073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,4,power_law_1.2,0.6070784091949463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,8,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,4,power_law_1.2,0.15229439735412598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,4,power_law_1.2,0.6642240047454834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,8,balanced,0.05101866523424784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,4,power_law_1.2,0.23198719024658204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,4,power_law_1.2,1.914771270751953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,8,balanced,0.04960533479849497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,4,power_law_1.2,0.3020096063613892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,8,balanced,0.05099200208981832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,4,power_law_1.2,0.46938238143920896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,8,balanced,0.04933333396911621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,4,power_law_1.2,0.6002367973327637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,8,balanced,0.05314666529496511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,4,power_law_1.2,0.6945792198181152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,8,balanced,0.05227733155091604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,4,power_law_1.2,1.1568320274353028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,8,balanced,0.05306133131186167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,4,power_law_1.2,1.8144256591796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,8,balanced,0.05716800192991892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,8,balanced,0.059578667084376015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,8,balanced,0.06486933430035909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,8,balanced,0.0736053337653478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,8,balanced,0.08407466610272725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,8,balanced,0.09673600395520528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,8,balanced,0.12828800082206726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,8,balanced,0.14990400274594626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,8,balanced,0.04176533222198486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,8,balanced,0.041877334316571556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,8,balanced,0.042954668402671814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,8,balanced,0.045221333702405296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,8,balanced,0.046053335070610046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,8,balanced,0.2019360065460205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,8,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,8,balanced,0.04500266909599304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,8,balanced,0.045253331462542214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,8,balanced,0.04510400195916494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,8,balanced,0.04515733321507772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,8,balanced,0.2566399971644084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,8,balanced,0.0451200008392334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,8,balanced,0.044794668753941856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,8,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,8,balanced,0.046495998899141945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,8,balanced,0.0489333321650823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,8,balanced,0.307370662689209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,8,balanced,0.051167999704678856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,8,balanced,0.053269331653912864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,8,balanced,0.05745600163936615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,8,balanced,0.058821335434913635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,8,balanced,0.0715946654478709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,8,balanced,0.45070934295654297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,8,balanced,0.08355733752250671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,8,balanced,0.11558933059374492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,8,balanced,0.1376426617304484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,8,balanced,0.18929066260655722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,8,balanced,0.2327359914779663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,8,balanced,0.8629546960194906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,8,balanced,0.32465600967407227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,8,balanced,0.417738676071167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,8,balanced,0.5054613351821899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,8,balanced,0.042266666889190674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,8,balanced,0.04141866664091746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,8,balanced,0.04186666508515676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,8,balanced,0.044794668753941856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,8,balanced,0.04526400069395701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,8,balanced,0.7775253454844157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,8,balanced,0.04508799811204275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,8,balanced,0.021429332594076794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,8,balanced,0.042021334171295166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,8,balanced,0.021685334543387096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,8,balanced,0.04499199986457825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,8,balanced,0.021984001000722248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,8,balanced,0.0451200008392334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,8,balanced,0.02179733415444692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,8,balanced,0.044693330923716225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,8,balanced,0.02141333371400833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,8,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,8,balanced,0.024325333535671234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,8,balanced,0.04381333291530609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,8,balanced,0.04684266448020935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,8,balanced,0.04484266539414724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,8,balanced,0.047050664822260536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,8,balanced,0.044581333796183266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,8,balanced,0.034629332522551216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,8,balanced,0.04722133278846741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,8,balanced,0.034304000437259674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,8,balanced,0.04914666712284088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,8,balanced,0.03639466563860575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,8,balanced,0.052906667192777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,8,balanced,0.03845866769552231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,8,balanced,0.054570664962132774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,8,balanced,0.037989333271980286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,8,balanced,0.0568800022204717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,8,balanced,0.03810133288304011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,8,balanced,0.03926933308442434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,8,balanced,0.0635040005048116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,8,balanced,0.03881600002447764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,8,balanced,1.5139946937561035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,8,balanced,0.07095466554164886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,8,balanced,0.04266666869322459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,8,balanced,0.045365333557128906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,8,balanced,0.09185600280761719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,8,balanced,0.048911998669306435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,8,balanced,0.10272533694903056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,8,balanced,0.06727466483910878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,8,balanced,0.07495999832948048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,8,balanced,0.13321600357691446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,8,balanced,0.09900266925493877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,8,balanced,0.12335466345151265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,8,balanced,0.16305599610010782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,8,balanced,0.16780267159144083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,8,balanced,0.21791466077168783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,8,balanced,0.2047306696573893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,8,balanced,0.27289066712061566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,8,balanced,0.2967519958813985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,8,balanced,0.38280534744262695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,8,balanced,0.33006399869918823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,8,balanced,0.464197317759196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,8,power_law_1.01,0.04694400131702423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,8,balanced,0.49836798508961994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,8,power_law_1.01,0.04739840030670166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,8,power_law_1.01,0.04714879989624023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,8,power_law_1.01,0.04904319941997528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,8,balanced,0.7229119936625162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,8,power_law_1.01,0.04894079864025116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,8,power_law_1.01,0.04890879988670349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,8,power_law_1.01,0.048172798752784726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,8,power_law_1.01,0.04822399914264679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,8,power_law_1.01,0.04805760085582733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,8,power_law_1.01,0.04750080108642578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,8,power_law_1.01,0.046374401450157164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,8,power_law_1.01,0.04783360064029694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,8,power_law_1.01,0.048793599009513855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,8,power_law_1.01,0.04859519898891449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,8,balanced,0.9487093289693197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,8,power_law_1.01,0.05480319857597351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,8,power_law_1.01,0.05470719933509827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,8,balanced,1.4173226356506348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,8,power_law_1.01,0.06225919723510742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,8,power_law_1.01,0.06959360241889953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,8,power_law_1.01,0.07640320062637329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,8,power_law_1.01,0.09913600087165833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,8,power_law_1.01,0.11478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,8,power_law_1.01,0.16088320016860963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,8,power_law_1.01,0.19304319620132446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,8,power_law_1.01,0.26136319637298583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,8,power_law_1.01,0.32423040866851804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,8,power_law_1.01,0.4764863967895508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,8,power_law_1.01,0.6185472011566162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,8,power_law_1.01,0.8020288467407226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,8,power_law_1.01,1.1977472305297852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,8,power_law_1.01,2.5290752410888673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,8,power_law_1.01,0.03950720131397247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,8,power_law_1.01,0.03943679928779602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,8,power_law_1.01,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,8,power_law_1.01,0.04169600009918213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,8,power_law_1.01,0.042387199401855466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,8,power_law_1.01,0.043007999658584595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,8,power_law_1.01,0.04268800020217896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,8,power_law_1.01,0.04232319891452789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,8,power_law_1.01,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,8,power_law_1.01,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,8,power_law_1.01,0.04286719858646393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,8,power_law_1.01,0.04359680116176605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,8,power_law_1.01,0.04508160054683685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,8,power_law_1.01,0.04665600061416626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,8,power_law_1.01,0.05070080161094666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,8,power_law_1.01,0.05237759947776795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,8,power_law_1.01,0.05678719878196716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,8,power_law_1.01,0.06238080263137817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,8,power_law_1.01,0.0754688024520874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,8,power_law_1.01,0.10180480480194092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,8,power_law_1.01,0.12271360158920289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,8,power_law_1.01,0.15990400314331055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,8,power_law_1.01,0.20817279815673828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,8,power_law_1.01,0.2727679967880249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,8,power_law_1.01,0.3673919916152954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,8,power_law_1.01,0.5163455963134765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,8,power_law_1.01,0.694700813293457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,8,power_law_1.01,0.8296704292297363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,8,power_law_1.01,1.352236843109131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,8,power_law_1.01,2.6710847854614257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,8,power_law_1.01,0.01963520050048828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,8,power_law_1.01,0.020614400506019592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,8,power_law_1.01,0.03989759981632233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,8,power_law_1.01,0.019750399887561797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,8,power_law_1.01,0.04028159976005554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,8,power_law_1.01,0.019891199469566346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,8,power_law_1.01,0.040678399801254275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,8,power_law_1.01,0.020032000541687012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,8,power_law_1.01,0.04193919897079468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,8,power_law_1.01,0.022745600342750548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,8,power_law_1.01,0.0420415997505188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,8,power_law_1.01,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,8,power_law_1.01,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,8,power_law_1.01,0.04615040123462677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,8,power_law_1.01,0.04179840087890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,8,power_law_1.01,0.03343360126018524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,8,power_law_1.01,0.041868799924850465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,8,power_law_1.01,0.03160319924354553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,8,power_law_1.01,0.0416128009557724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,8,power_law_1.01,0.033913600444793704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,8,power_law_1.01,0.037248000502586365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,8,power_law_1.01,0.036083200573921205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,8,power_law_1.01,0.03666560053825378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,8,power_law_1.01,0.038815999031066896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,8,power_law_1.01,0.037452799081802365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,8,power_law_1.01,0.040627199411392215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,8,power_law_1.01,0.04423680007457733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,8,power_law_1.2,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.01,0.0488319993019104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,8,power_law_1.2,0.04746879935264588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.01,0.06618880033493042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,8,power_law_1.2,0.047686401009559634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.01,0.07439360022544861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,8,power_law_1.2,0.04895359873771667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.01,0.0996288001537323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,8,power_law_1.2,0.04986239969730377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.01,0.12300159931182861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,8,power_law_1.2,0.04903680086135864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.01,0.16874879598617554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,8,power_law_1.2,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.01,0.20515201091766358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,8,power_law_1.2,0.04803839921951294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.01,0.2965183973312378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,8,power_law_1.2,0.04663679897785187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.01,0.38105599880218505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,8,power_law_1.2,0.047628799080848695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,8,power_law_1.01,0.46321921348571776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,8,power_law_1.2,0.047366398572921756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,8,power_law_1.01,0.7253119945526123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,8,power_law_1.2,0.048102399706840514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,8,power_law_1.01,1.4166975975036622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,8,power_law_1.2,0.048518401384353635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,8,power_law_1.2,0.05058559775352478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,8,power_law_1.2,0.057599997520446776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,8,power_law_1.2,0.054425597190856934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,8,power_law_1.2,0.061260801553726194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,8,power_law_1.2,0.06991999745368957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,8,power_law_1.2,0.07896959781646729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,8,power_law_1.2,0.10656640529632569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,8,power_law_1.2,0.11672320365905761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,8,power_law_1.2,0.1563007950782776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,8,power_law_1.2,0.19432320594787597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,8,power_law_1.2,0.28531200885772706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,8,power_law_1.2,0.3332223892211914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,8,power_law_1.2,0.5021567821502686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,8,power_law_1.2,0.6105279922485352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,8,power_law_1.2,0.7449024200439454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,8,power_law_1.2,1.2685183525085448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,8,power_law_1.2,2.521241569519043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,8,power_law_1.2,0.04010240137577057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,8,power_law_1.2,0.03969280123710632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,8,power_law_1.2,0.040575999021530154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,8,power_law_1.2,0.04243200123310089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,8,power_law_1.2,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,8,power_law_1.2,0.040934398770332336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,8,power_law_1.2,0.042822399735450746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,8,power_law_1.2,0.04178560078144074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,8,power_law_1.2,0.04247680008411407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,8,power_law_1.2,0.04095360040664673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,8,power_law_1.2,0.0433023989200592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,8,power_law_1.2,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,8,power_law_1.2,0.04672000110149384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,8,power_law_1.2,0.04423039853572845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,8,power_law_1.2,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,8,power_law_1.2,0.05236480236053467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,8,power_law_1.2,0.05687680244445801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,8,power_law_1.2,0.06321920156478882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,8,power_law_1.2,0.07210239768028259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,8,power_law_1.2,0.10042879581451417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,8,power_law_1.2,0.1244096040725708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,8,power_law_1.2,0.16663039922714235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,8,power_law_1.2,0.19996800422668456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,8,power_law_1.2,0.28837759494781495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,8,power_law_1.2,0.3422976016998291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,8,power_law_1.2,0.5367936134338379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,8,power_law_1.2,0.6843520164489746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,8,power_law_1.2,0.8700287818908692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,8,power_law_1.2,1.3617471694946288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,8,power_law_1.2,2.5570240020751953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,8,power_law_1.2,0.040454399585723874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,8,power_law_1.2,0.04118399918079376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,8,power_law_1.2,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,8,power_law_1.2,0.043699198961257936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,8,power_law_1.2,0.04188160002231598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,8,power_law_1.2,0.04121600091457367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,8,power_law_1.2,0.04252159893512726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,8,power_law_1.2,0.040863999724388124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,8,power_law_1.2,0.04191359877586365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,8,power_law_1.2,0.04154239892959595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,8,power_law_1.2,0.04087679982185364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,8,power_law_1.2,0.0462336003780365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,8,power_law_1.2,0.04376319944858551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,8,power_law_1.2,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,8,power_law_1.2,0.04911360144615173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,8,power_law_1.2,0.0506496012210846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,8,power_law_1.2,0.053907197713851926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,8,power_law_1.2,0.06159999966621399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.2,0.07203199863433837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,8,power_law_1.2,0.02021760046482086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.2,0.08748800158500672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,8,power_law_1.2,0.020883199572563172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.2,0.1034432053565979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,8,power_law_1.2,0.021279999613761903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.2,0.13258880376815796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,8,power_law_1.2,0.020908799767494202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.2,0.1600767970085144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,8,power_law_1.2,0.021510399878025055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.2,0.22099840641021729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,8,power_law_1.2,0.023897600173950196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.2,0.28152320384979246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,8,power_law_1.2,0.046726399660110475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.2,0.4067647933959961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,8,power_law_1.01,0.04053759872913361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,8,power_law_1.2,0.046777600049972536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.2,0.5271552085876465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,8,power_law_1.01,0.04192639887332916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,8,power_law_1.2,0.033670398592948916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,8,power_law_1.01,0.04440320134162903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,8,power_law_1.2,0.6370368003845215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,8,power_law_1.2,0.03506560027599335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,8,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,8,power_law_1.2,1.0164095878601074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,8,power_law_1.2,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,8,power_law_1.01,0.046374401450157164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,8,power_law_1.2,1.9203712463378906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,8,power_law_1.2,0.03834240138530731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,8,power_law_1.01,0.05073919892311096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,8,power_law_1.2,0.037241598963737486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,8,power_law_1.01,0.04971520006656647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,8,power_law_1.2,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,8,power_law_1.01,0.054201602935791016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,8,power_law_1.2,0.03954559862613678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,8,power_law_1.01,0.06371840238571166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,8,power_law_1.2,0.03844479918479919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.01,0.07118080258369446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,8,power_law_1.2,0.04222080111503601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.01,0.08718720078468323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,8,power_law_1.2,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.01,0.10053119659423829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.2,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.01,0.1261952042579651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.2,0.06760960221290588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.01,0.15479040145874023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.2,0.07534080147743225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.01,0.22001280784606933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.2,0.09945600032806397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.01,0.2746815919876099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.2,0.12339839935302735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.01,0.38485119342803953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.2,0.1685312032699585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.01,0.5204415798187256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.2,0.20456318855285643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.2,0.29631359577178956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.2,0.38046081066131593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,8,power_law_1.2,0.4637440204620361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,8,power_law_1.2,0.7245888233184814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,8,power_law_1.2,1.4185983657836914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,4,balanced,0.04427200059096018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,4,balanced,0.04532266656557719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,4,balanced,0.04749333361784617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,4,balanced,0.05068266888459524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,4,balanced,0.05153066913286845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,4,balanced,0.04980266590913137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,4,balanced,0.05055999755859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,4,balanced,0.049440001447995506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,4,balanced,0.04845866560935974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,4,balanced,0.04985600213209788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,4,balanced,0.049173335234324135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,4,balanced,0.048991998036702476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,4,balanced,0.05154666801293691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,4,balanced,0.04915200173854828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,4,balanced,0.05310399830341339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,4,balanced,0.05470933516820272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,4,balanced,0.05460800230503082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,4,balanced,0.06106133262316386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,4,balanced,0.06299200157324474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,4,balanced,0.07323200007279713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,4,balanced,0.08187200129032135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,4,balanced,0.09354133407274882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,4,balanced,0.10697600245475769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,4,balanced,0.14458666245142618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,4,balanced,0.1702453295389811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,4,balanced,0.23450666666030884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,4,balanced,0.037962667644023895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,4,balanced,0.03860266755024592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,4,balanced,0.04136000076929728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,4,balanced,0.2965386708577474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,4,balanced,0.0432586669921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,4,balanced,0.0425546665986379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,4,balanced,0.04450133442878723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,4,balanced,0.047168001532554626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,4,balanced,0.045114666223526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,4,balanced,0.3565066655476888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,4,balanced,0.04487466812133789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,4,balanced,0.0459146648645401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,4,balanced,0.04667200148105621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,4,balanced,0.04659733176231384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,4,balanced,0.0450133333603541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,4,balanced,0.04580800235271454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,4,balanced,0.051039998730023704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,4,balanced,0.5305013259251913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,4,balanced,0.05108266572157542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,4,balanced,0.05301866432030996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,4,balanced,0.060032000144322716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,4,balanced,0.06478933493296306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,4,balanced,0.07573333382606506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,4,balanced,0.037674665451049805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,4,balanced,0.09360000491142273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,4,balanced,0.03751466671625773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,4,balanced,0.04474133253097534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,4,balanced,0.12459199627240498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,4,balanced,0.044826666514078774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,4,balanced,0.04471466441949209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,4,balanced,0.150629331668218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,4,balanced,0.044735997915267944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,4,balanced,0.04484266539414724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,4,balanced,1.0140213171641033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,4,balanced,0.04911466439565023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,4,balanced,0.20526933670043945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,4,balanced,0.045381332437197365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,4,balanced,0.044346665342648826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,4,balanced,0.25170133511225384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,4,balanced,0.04520000020662943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,4,balanced,0.04458666841189066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,4,balanced,0.047450666626294456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,4,balanced,0.04709866642951965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,4,balanced,0.35603201389312744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,4,balanced,0.0491893341143926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,4,balanced,0.05129600067933401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,4,balanced,0.052629331747690834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,4,balanced,0.05727999905745188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,4,balanced,0.4599253336588542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,4,balanced,0.059402664502461754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,4,balanced,0.06843733290831248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,4,balanced,0.07930133243401845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,4,balanced,0.5672800143559774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,4,balanced,0.09986666838328044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,4,balanced,0.11458133657773335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,4,balanced,0.15109866857528687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,4,balanced,0.8691733678181967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,4,balanced,0.18441067139307657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,4,balanced,0.2563413381576538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,4,balanced,0.32042133808135986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,4,balanced,0.3889919916788737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,4,balanced,1.689893404642741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,8,power_law_1.01,0.6320831775665283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,4,balanced,0.597055991490682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,8,power_law_1.01,0.90829439163208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,8,power_law_1.01,1.8749824523925782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,4,balanced,1.1449546813964844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,4,balanced,0.02128533273935318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,4,balanced,0.02219199885924657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,4,balanced,0.02164799968401591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,4,balanced,0.022463999688625336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,4,balanced,0.022474666436513264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,4,balanced,0.021850667893886566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,4,balanced,0.024480000138282776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,4,balanced,0.025407999753952026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,4,balanced,0.05115733544031779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,4,balanced,0.04828799764315287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,4,balanced,0.04905599852403005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,4,balanced,0.03692800054947535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,4,balanced,0.03659733384847641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,4,balanced,0.038202665746212006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,4,balanced,0.03846933444341024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,4,balanced,0.03875733415285746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,4,balanced,0.042677332957585655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,4,balanced,0.045882667104403176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,4,balanced,0.04864533245563507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,4,balanced,0.05359466870625814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,4,balanced,0.06298133234182994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,4,balanced,0.08785600463549297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,4,balanced,0.09916266798973083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,4,balanced,0.13380266229311624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,4,balanced,0.16957332690556845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,4,balanced,0.23771200577418009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,4,power_law_1.01,0.04012160003185272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,4,power_law_1.01,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,4,power_law_1.01,0.04311679899692535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,4,balanced,0.29823466142018634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,4,power_law_1.01,0.03918719887733459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,4,power_law_1.01,0.047270399332046506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,4,power_law_1.01,0.04132480025291443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,4,power_law_1.01,0.04371840059757233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,4,power_law_1.01,0.042719998955726625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,4,power_law_1.01,0.048844799399375916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,4,power_law_1.01,0.042745599150657655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,4,power_law_1.01,0.04894720017910004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,4,power_law_1.01,0.04297600090503693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,4,balanced,0.368282675743103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,4,power_law_1.01,0.048556798696517946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,4,power_law_1.01,0.04361599981784821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,4,power_law_1.01,0.04910080134868622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,4,power_law_1.01,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,4,power_law_1.01,0.04867199957370758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,4,power_law_1.01,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,4,power_law_1.01,0.044921600818634035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,4,power_law_1.01,0.04950399994850159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,4,power_law_1.01,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,4,power_law_1.01,0.037484800815582274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,4,power_law_1.01,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,4,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,4,power_law_1.01,0.039027199149131775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,4,power_law_1.01,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,4,balanced,0.5682400067647299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,4,power_law_1.01,0.04609920084476471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,4,power_law_1.01,0.04317440092563629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,4,power_law_1.01,0.049446401000022885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,4,power_law_1.01,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,4,power_law_1.01,0.04289920032024384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,4,power_law_1.01,0.0523967981338501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,4,power_law_1.01,0.05236480236053467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,4,power_law_1.01,0.043654400110244754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,4,power_law_1.01,0.05185279846191406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,4,power_law_1.01,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,4,power_law_1.01,0.04425599873065948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,4,power_law_1.01,0.05559679865837097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,4,power_law_1.01,0.05607680082321167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,4,power_law_1.01,0.044012799859046936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,4,power_law_1.01,0.05848960280418396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,4,power_law_1.01,0.06963199973106385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,4,power_law_1.01,0.044275200366973876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,4,power_law_1.01,0.062015998363494876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,4,power_law_1.01,0.07523199915885925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,4,power_law_1.01,0.043289598822593686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,4,power_law_1.01,0.07497599720954895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,4,power_law_1.01,0.09217919707298279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,4,power_law_1.01,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,4,power_law_1.01,0.07454720139503479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,4,power_law_1.01,0.12250239849090576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,4,power_law_1.01,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,4,power_law_1.01,0.09628159999847412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,4,power_law_1.01,0.15960320234298705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,4,power_law_1.01,0.044710400700569156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,4,power_law_1.01,0.10718719959259033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,4,power_law_1.01,0.04565759897232056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,4,power_law_1.01,0.17802239656448365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,4,power_law_1.01,0.14659839868545532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,4,balanced,1.0949973265329997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,4,power_law_1.01,0.05208960175514221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,4,power_law_1.01,0.29351038932800294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,4,power_law_1.01,0.16167039871215821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,4,power_law_1.01,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,4,power_law_1.01,0.37696640491485595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,4,power_law_1.01,0.22660479545593262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,4,power_law_1.01,0.05559039711952209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,4,power_law_1.01,0.5109951972961426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,4,power_law_1.01,0.30305919647216795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,4,power_law_1.01,0.06458240151405334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,4,power_law_1.01,0.7330495834350585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,4,power_law_1.01,0.4331711769104004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.01,0.07352960109710693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,4,power_law_1.01,0.8642368316650391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,4,power_law_1.01,0.6314559936523437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.01,0.08997120261192322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,4,power_law_1.01,0.7366784095764161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,4,power_law_1.01,1.3841792106628419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.01,0.09941759705543518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,4,power_law_1.01,1.0399744033813476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,4,power_law_1.01,2.8286272048950196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.01,0.13864959478378297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,4,power_law_1.01,2.23819522857666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.01,0.1525887966156006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.01,0.23274240493774415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.01,0.324563193321228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.01,0.4711935997009277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.01,0.4641727924346924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,4,power_law_1.01,0.7532479763031006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,4,power_law_1.01,1.0795071601867676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,4,power_law_1.01,2.1612672805786133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,4,power_law_1.01,0.01849599927663803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,4,power_law_1.01,0.019622400403022766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,4,power_law_1.01,0.020153599977493285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,4,power_law_1.01,0.021011200547218323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,4,power_law_1.01,0.01982080042362213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,4,power_law_1.01,0.020710399746894835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,4,power_law_1.01,0.02335360050201416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,4,power_law_1.01,0.023027199506759643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,4,power_law_1.2,0.04599039852619171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,4,power_law_1.01,0.04850560128688812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,4,power_law_1.2,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,4,power_law_1.2,0.03893119990825653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,4,power_law_1.01,0.04672000110149384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,4,power_law_1.2,0.04726400077342987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,4,power_law_1.2,0.038790398836135866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,4,power_law_1.01,0.04694400131702423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,4,power_law_1.2,0.04744960069656372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,4,power_law_1.2,0.03944959938526153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,4,power_law_1.01,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,4,power_law_1.2,0.04772480130195618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,4,power_law_1.01,0.03431679904460907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,4,power_law_1.2,0.041305598616600034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,4,power_law_1.2,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,4,power_law_1.2,0.04270080029964447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,4,power_law_1.01,0.0350847989320755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,4,power_law_1.2,0.049267199635505673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,4,power_law_1.2,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,4,power_law_1.2,0.043372800946235655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,4,power_law_1.2,0.048691201210021975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,4,power_law_1.2,0.03824639916419983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,4,power_law_1.2,0.043808001279830935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,4,power_law_1.2,0.049439999461174014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,4,power_law_1.2,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,4,power_law_1.2,0.03835520148277283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,4,power_law_1.2,0.04959999918937683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,4,power_law_1.2,0.04413439929485321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,4,power_law_1.2,0.04950399994850159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,4,power_law_1.2,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,4,power_law_1.2,0.04497919976711273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,4,power_law_1.2,0.05000960230827332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,4,power_law_1.2,0.04279040098190308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,4,power_law_1.2,0.045587199926376346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,4,power_law_1.2,0.052908802032470705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,4,power_law_1.2,0.04287999868392944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,4,power_law_1.2,0.04620800018310547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,4,power_law_1.2,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,4,power_law_1.2,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,4,power_law_1.2,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,4,power_law_1.2,0.05891839861869812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,4,power_law_1.2,0.04471679925918579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,4,power_law_1.2,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,4,power_law_1.2,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,4,power_law_1.2,0.04534400105476379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,4,power_law_1.2,0.05616000294685364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,4,power_law_1.2,0.06151679754257202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,4,power_law_1.2,0.0455487996339798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,4,power_law_1.2,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,4,power_law_1.2,0.07415680289268493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,4,power_law_1.2,0.0436928004026413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,4,power_law_1.2,0.06917120218276977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,4,power_law_1.2,0.07942399978637696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,4,power_law_1.2,0.0843775987625122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,4,power_law_1.2,0.04439040124416351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,4,power_law_1.2,0.1020799994468689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,4,power_law_1.2,0.10363520383834839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,4,power_law_1.2,0.12319359779357911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,4,power_law_1.2,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,4,power_law_1.2,0.12783360481262207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,4,power_law_1.2,0.14589439630508422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,4,power_law_1.2,0.04776319861412048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,4,power_law_1.2,0.17337599992752076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,4,power_law_1.2,0.182259202003479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,4,power_law_1.2,0.04780159890651703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,4,power_law_1.2,0.20977280139923096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,4,power_law_1.2,0.2644223928451538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,4,power_law_1.2,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,4,power_law_1.2,0.31045119762420653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,4,power_law_1.2,0.32928640842437745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,4,power_law_1.2,0.0567359983921051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,4,power_law_1.2,0.38648319244384766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,4,power_law_1.2,0.4598976135253906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,4,power_law_1.2,0.05960320234298706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,4,power_law_1.2,0.5413375854492187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,4,power_law_1.2,0.4966591835021973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,4,power_law_1.2,0.06509439945220948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,4,power_law_1.2,0.6832640171051025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,4,power_law_1.2,0.7217984199523926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.2,0.0782912015914917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,4,power_law_1.2,0.8917695999145507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,4,power_law_1.2,1.337600040435791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.2,0.08873599767684937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,4,power_law_1.2,1.6393152236938477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,4,power_law_1.2,2.4362688064575195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.2,0.10915199518203736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,4,power_law_1.2,3.106879997253418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.2,0.14245760440826416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.2,0.19029120206832886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.2,0.2393791913986206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.2,0.3387840032577515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.2,0.4668799877166748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.2,0.5953023910522461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,4,power_law_1.2,0.7757760047912597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,4,power_law_1.2,1.2265600204467773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,4,power_law_1.2,2.378816032409668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,4,power_law_1.2,0.01879040002822876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,4,power_law_1.2,0.01982080042362213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,4,power_law_1.2,0.019776000082492827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,4,power_law_1.2,0.01976960003376007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,4,power_law_1.2,0.020319999754428865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,4,power_law_1.2,0.020102399587631225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,4,power_law_1.2,0.021855999529361726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,4,power_law_1.2,0.023263999819755556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,4,power_law_1.2,0.04655359983444214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,4,power_law_1.2,0.046323201060295104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,4,power_law_1.2,0.04659200012683869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,4,power_law_1.2,0.034092798829078674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,4,power_law_1.2,0.03425920009613037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,4,power_law_1.2,0.03518719971179962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,4,power_law_1.2,0.03665919899940491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,4,power_law_1.2,0.03702400028705597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,4,power_law_1.2,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,4,power_law_1.2,0.04252159893512726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.2,0.047884801030159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.2,0.05374720096588135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.2,0.06174719929695129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.2,0.08572160005569458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.2,0.09895679950714112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.2,0.13322880268096923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.2,0.16593919992446898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.2,0.23779840469360353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.2,0.2972415924072266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,4,power_law_1.2,0.36726400852203367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,4,power_law_1.01,0.03639039993286133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,4,power_law_1.2,0.5683904170989991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,4,power_law_1.01,0.038412800431251524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,4,power_law_1.01,0.04084480106830597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,4,power_law_1.2,1.0928128242492676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,4,power_law_1.01,0.04631040096282959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.01,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,128,balanced,0.08338133494059245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.01,0.05447040200233459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,128,balanced,0.08549867073694865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.01,0.062438398599624634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,128,balanced,0.08500267068545024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.01,0.0859391987323761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,128,balanced,0.0881173312664032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.01,0.09839360117912292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,128,balanced,0.08725866675376892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.01,0.1347264051437378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,128,balanced,0.10016533732414246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.01,0.1692031979560852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,128,balanced,0.1009173293908437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,128,balanced,0.10215466221173604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.01,0.237990403175354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,128,balanced,0.0997279981772105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.01,0.2983936071395874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,128,balanced,0.06085333228111267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,128,balanced,0.1042133371035258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,128,balanced,0.06469866633415222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,4,power_law_1.01,0.36828160285949707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,128,balanced,0.06325866778691609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,128,balanced,0.10614933570226033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,4,power_law_1.01,0.5697919845581054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,128,balanced,0.06369600196679433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,128,balanced,0.10942400495211284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,4,power_law_1.01,1.0953344345092773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,128,balanced,0.06785066425800323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,128,balanced,0.11940800150235494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,128,balanced,0.07414933542410533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,128,balanced,0.10289067029953003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,128,balanced,0.0744053324063619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,128,balanced,0.07551999886830647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,128,balanced,0.10912000139554341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,128,balanced,0.07813333471616109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,128,balanced,0.11425066987673442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,128,balanced,0.07638933261235555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,128,balanced,0.11930132905642192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,128,balanced,0.07971733311812083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,128,balanced,0.07930666704972585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,128,balanced,0.11828266580899556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,128,balanced,0.07764799892902374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,128,balanced,0.1269653340180715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,128,balanced,0.086517333984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,128,balanced,0.08348799745241801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,128,balanced,0.12446932991345723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,128,balanced,0.08706667025883992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,128,balanced,0.096261332432429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,128,balanced,0.1376106639703115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,128,balanced,0.10077866911888123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,128,balanced,0.10974400242169698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,128,balanced,0.15442132949829102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,128,balanced,0.12819199760754904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,128,balanced,0.14693333705266318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,128,balanced,0.17368000745773315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,128,balanced,0.18107734123865762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,128,balanced,0.2258666753768921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,128,balanced,0.21442667643229166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,128,balanced,0.2879573305447896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,128,balanced,0.25643734137217206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,128,balanced,0.35411731402079266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,128,balanced,0.36028798421223956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,128,balanced,0.499781330426534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,128,balanced,0.4357279936472575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,128,balanced,0.6359573205312093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,128,balanced,0.7771360079447428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,128,balanced,0.5286879936854044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,128,balanced,0.06172800064086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,128,balanced,0.0651093324025472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,128,balanced,0.06400000055631001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,128,balanced,0.0629013329744339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,128,balanced,0.06764266888300578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,128,balanced,0.0747626672188441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,128,balanced,0.07749333480993907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,128,balanced,0.07560533285140991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,128,balanced,0.07929599781831105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,128,balanced,0.07649066547552745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,128,balanced,0.07579733431339264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,128,balanced,0.07741333544254303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,128,balanced,0.08003200093905131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,128,balanced,0.07846400141716003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,128,balanced,0.0831573357184728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,128,balanced,1.19487468401591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,128,balanced,0.023717333873112995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,128,balanced,0.08529599507649739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,128,balanced,0.02685333291689555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,128,balanced,0.08705600102742513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,128,balanced,0.7968959808349609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,128,balanced,0.025653332471847534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,128,balanced,0.028389332195123036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,128,balanced,0.092549333969752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,128,balanced,0.03401066611210505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,128,balanced,0.05526400109132131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,128,balanced,0.0942186713218689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,128,balanced,0.05273066461086273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,128,balanced,0.05305600166320801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,128,balanced,0.10804800192515056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,128,balanced,0.05299733579158783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,128,balanced,0.05251200000445048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,128,balanced,0.12300266822179158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,128,balanced,0.051557332277297974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,128,balanced,0.04924799998601278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,128,balanced,0.04629333317279816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,128,balanced,0.14642666776974997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,128,balanced,0.044122666120529175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,128,balanced,0.060906668504079185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,128,balanced,0.06190933287143707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,128,balanced,0.05052266518274943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,128,balanced,0.1718133290608724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,128,balanced,0.07036266724268596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,128,balanced,0.07740800082683563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,128,balanced,0.10170132915178935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,128,balanced,0.22971200942993164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,128,balanced,0.12229333321253459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,128,balanced,0.17101333538691202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,128,balanced,0.2697333296140035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,128,balanced,0.20843732357025146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,128,balanced,0.37276268005371094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,128,balanced,0.30721066395441693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,128,balanced,2.344655990600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,128,balanced,0.40081600348154706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,128,balanced,1.5307092666625977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,128,balanced,0.472053329149882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,128,balanced,0.5912160078684489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,128,balanced,0.5713653167088827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,128,balanced,0.7799200216929117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,128,balanced,0.9654346307118734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,128,balanced,0.8715999921162924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,128,balanced,1.5248746871948242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,128,balanced,1.703887939453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,128,power_law_1.01,0.09189760088920593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,128,power_law_1.01,0.09694079756736755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,128,power_law_1.01,0.09513599872589111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,128,power_law_1.01,0.09488000273704529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,128,power_law_1.01,0.09956480264663696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,128,power_law_1.01,0.07390720248222352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,128,power_law_1.01,0.10302079916000366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,128,power_law_1.01,0.075135999917984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,128,power_law_1.01,0.10403200387954711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,128,power_law_1.01,0.06748160123825073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,128,balanced,3.003077189127604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,128,power_law_1.01,0.10025600194931031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,128,power_law_1.01,0.06876159906387329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,128,power_law_1.01,0.10403200387954711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,128,power_law_1.01,0.07199360132217407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,128,power_law_1.01,0.10396159887313842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,128,power_law_1.01,0.07605119943618774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,128,power_law_1.01,0.10334080457687378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,128,power_law_1.01,0.07779840230941773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,128,power_law_1.01,0.09946240186691284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,128,power_law_1.01,0.07549440264701843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,128,power_law_1.01,0.10088319778442383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,128,power_law_1.01,0.07493759989738465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,128,power_law_1.01,0.10112639665603637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,128,power_law_1.01,0.07644799947738648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,128,power_law_1.01,0.07561600208282471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,128,power_law_1.01,0.10727039575576783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,128,power_law_1.01,0.07794560194015503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,128,power_law_1.01,0.10152319669723511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,128,power_law_1.01,0.0763264000415802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,128,power_law_1.01,0.10166399478912354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,128,power_law_1.01,0.08280959725379944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,128,power_law_1.01,0.1218176007270813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,128,power_law_1.01,0.08506240248680115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,128,power_law_1.01,0.12190079689025879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,128,power_law_1.01,0.08756480216979981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,128,power_law_1.01,0.16556799411773682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,128,power_law_1.01,0.09065600037574768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,128,power_law_1.01,0.1772096037864685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,128,power_law_1.01,0.1055232048034668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,128,power_law_1.01,0.20817921161651612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,128,power_law_1.01,0.11560959815979004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,128,power_law_1.01,0.2625920057296753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,128,power_law_1.01,0.1480064034461975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,128,power_law_1.01,0.33177599906921384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,128,power_law_1.01,0.17149440050125123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,128,power_law_1.01,0.4483839988708496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,128,power_law_1.01,0.22014079093933106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,128,power_law_1.01,0.6003776073455811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,128,power_law_1.01,0.2658368110656738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,128,power_law_1.01,0.8309184074401855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,128,power_law_1.01,0.3819648027420044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,128,power_law_1.01,1.0297216415405273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,128,power_law_1.01,0.4675903797149658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,128,power_law_1.01,1.5317888259887695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,128,power_law_1.01,0.6731200218200684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,128,power_law_1.01,3.11657600402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,128,power_law_1.01,0.8569855690002441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,128,power_law_1.01,1.0864959716796876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,128,power_law_1.01,1.7766016006469727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,128,power_law_1.01,3.287628936767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,128,power_law_1.01,0.07367680072784424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,128,power_law_1.01,0.07299200296401978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,128,power_law_1.01,0.0691648006439209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,128,power_law_1.01,0.07846400141716003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,128,power_law_1.01,0.07348480224609374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,128,power_law_1.01,0.07438719868659974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,128,power_law_1.01,0.07371519804000855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,128,power_law_1.01,0.07404800057411194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,128,power_law_1.01,0.022540800273418427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,128,power_law_1.01,0.0738752007484436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,128,power_law_1.01,0.07249280214309692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,128,power_law_1.01,0.023897600173950196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,128,power_law_1.01,0.07383040189743043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,128,power_law_1.01,0.022899200022220612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,128,power_law_1.01,0.0754751980304718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,128,power_law_1.01,0.025702399015426636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,128,power_law_1.01,0.07656959891319275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,128,power_law_1.01,0.033497598767280576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,128,power_law_1.01,0.07738879919052125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,128,power_law_1.01,0.052883201837539674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,128,power_law_1.01,0.08072320222854615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,128,power_law_1.01,0.05111680030822754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,128,power_law_1.01,0.07979519963264466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,128,power_law_1.01,0.05320320129394531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,128,power_law_1.01,0.08181759715080261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,128,power_law_1.01,0.05231999754905701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,128,power_law_1.01,0.0998848021030426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,128,power_law_1.01,0.05198079943656921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.01,0.10410239696502685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,128,power_law_1.01,0.0522816002368927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.01,0.12670079469680787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,128,power_law_1.01,0.05242239832878113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.01,0.14496639966964722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,128,power_law_1.01,0.050374400615692136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.01,0.18109439611434935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,128,power_law_1.01,0.04713599979877472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.01,0.23474559783935547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,128,power_law_1.01,0.04686079919338226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,128,power_law_1.01,0.044486400485038755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.01,0.29385600090026853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,128,power_law_1.01,0.05333120226860046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.01,0.3764096021652222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,128,power_law_1.2,0.09831680059432983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,128,power_law_1.01,0.05699840188026428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.01,0.5631103992462159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,128,power_law_1.2,0.09093760251998902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.01,0.07245439887046815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.01,0.6954239845275879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,128,power_law_1.2,0.08758400082588196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.01,0.08520960211753845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,128,power_law_1.01,0.8317760467529297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,128,power_law_1.2,0.09889919757843017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.01,0.10101759433746338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,128,power_law_1.01,1.2592703819274902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,128,power_law_1.2,0.10051840543746948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.01,0.13627519607543945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,128,power_law_1.2,0.10157439708709717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.01,0.16717439889907837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,128,power_law_1.01,2.7742080688476562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,128,power_law_1.2,0.10587519407272339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.01,0.23523199558258057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,128,power_law_1.2,0.07278079986572265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,128,power_law_1.2,0.1028864026069641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.01,0.30580480098724366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,128,power_law_1.2,0.07683839797973632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,128,power_law_1.2,0.09946879744529724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.01,0.4275775909423828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,128,power_law_1.2,0.06725760102272034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,128,power_law_1.2,0.10215679407119752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.01,0.6014463901519775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,128,power_law_1.2,0.06815360188484192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,128,power_law_1.2,0.09815040230751038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,128,power_law_1.2,0.10210560560226441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,128,power_law_1.2,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,128,power_law_1.01,0.6951551914215088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,128,power_law_1.2,0.10332800149917602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,128,power_law_1.2,0.07362560033798218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,128,power_law_1.01,1.128486442565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,128,power_law_1.2,0.09878399968147278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,128,power_law_1.2,0.07255039811134338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,128,power_law_1.2,0.10364799499511719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,128,power_law_1.2,0.07400320172309875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,128,power_law_1.01,2.2398143768310548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,128,power_law_1.2,0.07719680070877075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,128,power_law_1.2,0.10201599597930908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,128,power_law_1.2,0.11764479875564575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,128,power_law_1.2,0.08051199913024902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,128,power_law_1.2,0.12877440452575684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,128,power_law_1.2,0.07921280264854431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,128,power_law_1.2,0.12877440452575684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,128,power_law_1.2,0.07923840284347534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,128,power_law_1.2,0.17322239875793458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,128,power_law_1.2,0.0834559977054596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,128,power_law_1.2,0.19953919649124147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,128,power_law_1.2,0.08305919766426087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,128,power_law_1.2,0.08291199803352356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,128,power_law_1.2,0.26791040897369384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,128,power_law_1.2,0.089273601770401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,128,power_law_1.2,0.37397758960723876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,128,power_law_1.2,0.09678720235824585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,128,power_law_1.2,0.4369984149932861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,128,power_law_1.2,0.10848000049591064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,128,power_law_1.2,0.5808127880096435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,128,power_law_1.2,0.11622400283813476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,128,power_law_1.2,0.8319231986999511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,128,power_law_1.2,0.1541759967803955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,128,power_law_1.2,1.1447487831115724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,128,power_law_1.2,0.17824000120162964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,128,power_law_1.2,1.5684543609619142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,128,power_law_1.2,0.23815679550170898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,128,power_law_1.2,2.819071960449219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,128,power_law_1.2,0.3184128046035767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,128,power_law_1.2,0.4415872097015381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,128,power_law_1.2,6.174399948120117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,128,power_law_1.2,0.5620416164398193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,128,power_law_1.2,0.8488639831542969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,128,power_law_1.2,1.0121855735778809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,128,power_law_1.2,1.3321855545043946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,128,power_law_1.2,2.1719039916992187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,128,power_law_1.2,4.938304138183594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,128,power_law_1.2,0.07154560089111328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,128,power_law_1.2,0.0728384017944336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,128,power_law_1.2,0.06324480175971985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,128,power_law_1.2,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,128,power_law_1.2,0.07125120162963867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,128,power_law_1.2,0.07293440103530884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,128,power_law_1.2,0.02152319997549057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,128,power_law_1.2,0.07409279942512512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,128,power_law_1.2,0.022944000363349915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,128,power_law_1.2,0.07391999959945679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,128,power_law_1.2,0.022835199534893037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,128,power_law_1.2,0.07649919986724854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,128,power_law_1.2,0.02592639923095703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,128,power_law_1.2,0.07385600209236146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,128,power_law_1.2,0.03344640135765076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,128,power_law_1.2,0.07564160227775574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,128,power_law_1.2,0.052883201837539674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,128,power_law_1.2,0.07451519966125489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,128,power_law_1.2,0.050329601764678954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,128,power_law_1.2,0.07486079931259156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,128,power_law_1.2,0.05081599950790405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,128,power_law_1.2,0.051148802042007446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,128,power_law_1.2,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,128,power_law_1.2,0.05070080161094666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,128,power_law_1.2,0.07896959781646729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,128,power_law_1.2,0.05090559720993042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,128,power_law_1.2,0.08085119724273682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,128,power_law_1.2,0.05137280225753784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,128,power_law_1.2,0.08830080032348633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,128,power_law_1.2,0.04966399967670441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,64,balanced,0.06426666676998138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,128,power_law_1.2,0.10598399639129638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,64,balanced,0.06592000027497609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,128,power_law_1.2,0.048153600096702574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.2,0.11004799604415894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,64,balanced,0.06645866731802623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,128,power_law_1.2,0.0434112012386322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.2,0.13257600069046022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,64,balanced,0.0674720009167989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,64,balanced,0.0784746656815211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,128,power_law_1.2,0.047865599393844604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.2,0.15847040414810182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,64,balanced,0.0962720016638438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,128,power_law_1.2,0.05975679755210876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.2,0.20139520168304442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,64,balanced,0.0988159974416097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,128,power_law_1.2,0.058822399377822875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,64,balanced,0.09161600470542908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.2,0.2515007972717285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,64,balanced,0.1039466659228007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.2,0.07729920148849487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.2,0.36620159149169923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,64,balanced,0.10194133718808492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.2,0.08867200016975403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.2,0.454969596862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,64,balanced,0.10698133707046509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.2,0.10661120414733886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,64,balanced,0.10321066776911418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.2,0.6794432163238525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,64,balanced,0.10192533334096272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.2,0.1383296012878418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.2,0.881772804260254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,64,balanced,0.10141332944234212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.2,0.1734976053237915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,128,power_law_1.2,1.14268798828125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,64,balanced,0.10763200124104817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.2,0.24967679977416993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,64,balanced,0.11216533184051514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,128,power_law_1.2,1.836511993408203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,64,balanced,0.05410666763782501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.2,0.33641600608825684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,64,balanced,0.10979732871055603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,64,balanced,0.05585066477457682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.2,0.44875521659851075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,128,power_law_1.2,3.381407928466797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,64,balanced,0.05805333455403646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,64,balanced,0.1144426663716634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,64,balanced,0.058933332562446594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.2,0.633516788482666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,64,balanced,0.06601066887378693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,64,balanced,0.11708266536394756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,64,balanced,0.07512533167997996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,128,power_law_1.2,0.7240511894226074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,64,balanced,0.07524266839027405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,64,balanced,0.12467199563980103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,128,power_law_1.2,1.3423040390014649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,64,balanced,0.07663999994595845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,64,balanced,0.0761653333902359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,64,balanced,0.13427733381589255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,128,power_law_1.2,2.716556739807129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,64,balanced,0.07397333284219106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,64,balanced,0.07619200150171916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,64,balanced,0.07825066645940144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,64,balanced,0.15597333510716757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,64,balanced,0.081535999973615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,64,balanced,0.08297599852085114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,64,balanced,0.08553600311279297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,64,balanced,0.1772480010986328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,64,balanced,0.09012266993522644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,64,balanced,0.0938933293024699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,64,balanced,0.10314666231473286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,64,balanced,0.2209226687749227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,64,balanced,0.11179733276367188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,64,balanced,0.13025599718093872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,64,balanced,0.2593066692352295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,64,balanced,0.1460586686929067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,64,balanced,0.1846933364868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,64,balanced,0.2174773414929708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,64,balanced,0.3621813456217448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,64,balanced,0.28965334097544354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,64,balanced,0.443994681040446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,64,balanced,0.35807998975118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,64,balanced,0.5092320044835409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,64,balanced,0.5469119946161906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,64,balanced,0.6455466747283936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,64,balanced,0.8112213611602783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,64,balanced,0.7939679622650146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,64,balanced,0.056202664971351624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,64,balanced,0.059279998143514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,64,balanced,0.056554665168126426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,64,balanced,0.058789332707722984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,64,balanced,0.06930666665236156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,64,balanced,0.07702933251857758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,64,balanced,0.07319466769695282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,64,balanced,0.07938666641712189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,64,balanced,0.07379733522733052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,64,balanced,0.07292800148328145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,64,balanced,0.023957334458827972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,64,balanced,0.024634666740894318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,64,balanced,0.0774293343226115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,64,balanced,0.025829332570234936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,64,balanced,0.07872533301512401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,64,balanced,0.026234666506449383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,64,balanced,0.0776800016562144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,64,balanced,0.026602665583292644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,64,balanced,0.08185066779454549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,64,balanced,0.03205333401759466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,64,balanced,0.08349866668383281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,64,balanced,0.055349335074424744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,64,balanced,0.056976000467936196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,64,balanced,0.08514133095741272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,64,balanced,0.05331199864546458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,64,balanced,0.08853866656621297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,64,balanced,0.053157334526379905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,64,balanced,1.216981331507365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,64,balanced,0.0934399962425232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,64,balanced,0.05312533179918925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,64,balanced,0.05109333495299021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,64,balanced,0.10207999746004741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,64,balanced,0.048751999934514366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,64,balanced,0.04651733239491781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,64,balanced,0.10998933513959248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,64,balanced,0.06286400059858958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,64,balanced,1.5600639979044597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,64,balanced,0.06438399851322174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,64,balanced,0.12359467148780823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,64,balanced,0.05258133510748545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,64,balanced,0.0757013310988744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,64,balanced,0.14877866705258688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,64,balanced,0.08238400022188823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,64,balanced,0.10788800319035848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,64,balanced,0.1734666625658671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,64,balanced,0.13060800234476724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,64,balanced,0.18257067600886026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,64,balanced,0.2304426630338033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,64,balanced,0.2225546638170878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,64,balanced,0.2757973273595174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,64,balanced,0.32421332597732544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,64,balanced,0.4238293170928955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,64,balanced,0.3814186652501424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,64,balanced,0.6312426726023356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,64,balanced,2.389450709025065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,64,balanced,0.4836159944534302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,64,balanced,0.8379573027292887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,64,balanced,0.5888959964116415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,64,balanced,1.0441493193308513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,64,balanced,0.8967040379842123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,64,balanced,1.6490772565205891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,64,power_law_1.01,0.09047039747238159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,64,power_law_1.01,0.08819199800491333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,64,power_law_1.01,0.08727040290832519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,64,power_law_1.01,0.08406400084495544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,64,power_law_1.01,0.08994560241699219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,64,power_law_1.01,0.0880128026008606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,64,power_law_1.01,0.09690880179405212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,64,power_law_1.01,0.09630079865455628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,64,power_law_1.01,0.09616640210151672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,64,power_law_1.01,0.09715200066566468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,64,power_law_1.01,0.09774079918861389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,64,power_law_1.01,0.10016640424728393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,64,power_law_1.01,0.0967296004295349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,64,power_law_1.01,0.09852160215377807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,64,power_law_1.01,0.10160640478134156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,64,power_law_1.01,0.10381439924240113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,64,power_law_1.01,0.10284160375595093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,64,power_law_1.01,0.1134719967842102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,64,power_law_1.01,0.12280319929122925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,64,power_law_1.01,0.14380799531936644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,64,balanced,1.7519307136535645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,64,power_law_1.01,0.1684864044189453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,64,power_law_1.01,0.20276479721069335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,64,power_law_1.01,0.22430720329284667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,64,power_law_1.01,0.31480960845947265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,64,power_law_1.01,0.36472320556640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,64,balanced,3.2608105341593423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,64,power_law_1.01,0.5922431945800781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,64,power_law_1.01,0.726259183883667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,64,power_law_1.01,0.9219712257385254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,64,power_law_1.01,1.3230463981628418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,64,power_law_1.01,0.07128959894180298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,64,power_law_1.01,3.03570556640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,64,power_law_1.01,0.06527360081672669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,64,power_law_1.01,0.06471040248870849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,64,power_law_1.01,0.06414080262184144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,64,power_law_1.01,0.06635519862174988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,64,power_law_1.01,0.07134720087051391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,64,power_law_1.01,0.06968320012092591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,64,power_law_1.01,0.07646080255508422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,64,power_law_1.01,0.07134720087051391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,64,power_law_1.01,0.0743552029132843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,64,power_law_1.01,0.07741439938545228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,64,power_law_1.01,0.07534080147743225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,64,power_law_1.01,0.07626879811286927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,64,power_law_1.01,0.07720320224761963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,64,power_law_1.01,0.08561279773712158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,64,power_law_1.01,0.08567039966583252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,64,power_law_1.01,0.09377279877662659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,64,power_law_1.01,0.10373760461807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,64,power_law_1.01,0.11407359838485717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,64,power_law_1.01,0.13724160194396973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,64,power_law_1.01,0.16346880197525024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,64,power_law_1.01,0.21575040817260743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,64,power_law_1.01,0.245580792427063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,64,power_law_1.01,0.34339840412139894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,64,power_law_1.01,0.4419519901275635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,64,power_law_1.01,0.6163712024688721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,64,power_law_1.01,0.8211584091186523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,64,power_law_1.01,1.019859218597412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,64,power_law_1.01,1.555622386932373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,64,power_law_1.01,3.1549823760986326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,64,power_law_1.01,0.06920959949493408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,64,power_law_1.01,0.06935679912567139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,64,power_law_1.01,0.07238399982452393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,64,power_law_1.01,0.0640447974205017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,64,power_law_1.01,0.06954240202903747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,64,power_law_1.01,0.02227199971675873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,64,power_law_1.01,0.06584320068359376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,64,power_law_1.01,0.023737600445747374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,64,power_law_1.01,0.07240319848060608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,64,power_law_1.2,0.09144960045814514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,64,power_law_1.01,0.022572800517082214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,64,power_law_1.01,0.07637760043144226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,64,power_law_1.2,0.09233919978141784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,64,power_law_1.01,0.02343039959669113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,64,power_law_1.01,0.07027199864387512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,64,power_law_1.2,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,64,power_law_1.01,0.02447360008955002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,64,power_law_1.01,0.07357439994812012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,64,power_law_1.2,0.08010240197181702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,64,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,64,power_law_1.01,0.0725823998451233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,64,power_law_1.2,0.08512639999389648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,64,power_law_1.01,0.05440000295639038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,64,power_law_1.01,0.0759935975074768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,64,power_law_1.2,0.08851839900016785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,64,power_law_1.01,0.05456640124320984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,64,power_law_1.2,0.0915391981601715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,64,power_law_1.01,0.07674880027770996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,64,power_law_1.01,0.05329279899597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,64,power_law_1.2,0.09544960260391236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,64,power_law_1.01,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,64,power_law_1.01,0.05370879769325256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,64,power_law_1.2,0.09273599982261657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,64,power_law_1.01,0.08021119832992554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,64,power_law_1.01,0.05334399938583374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,64,power_law_1.2,0.09681919813156128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,64,power_law_1.01,0.08257920145988465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,64,power_law_1.01,0.05306239724159241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,64,power_law_1.2,0.09651200175285339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,64,power_law_1.01,0.08330879807472229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,64,power_law_1.01,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,64,power_law_1.2,0.09498879909515381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,64,power_law_1.01,0.09781119823455811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,64,power_law_1.01,0.05041279792785645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,64,power_law_1.2,0.09825279712677001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.01,0.10943360328674316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,64,power_law_1.01,0.05047680139541626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,64,power_law_1.2,0.09675520062446594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.01,0.12714240550994874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,64,power_law_1.01,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,64,power_law_1.2,0.10451200008392333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.01,0.14351999759674072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,64,power_law_1.01,0.05516800284385681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,64,power_law_1.2,0.10362240076065063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.01,0.18024319410324097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,64,power_law_1.01,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,64,power_law_1.2,0.11230720281600952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.01,0.21271679401397706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.01,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,64,power_law_1.2,0.11291520595550537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.01,0.08172799944877625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.01,0.3001983880996704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,64,power_law_1.2,0.12776960134506227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.01,0.09027199745178223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.01,0.3545664072036743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,64,power_law_1.2,0.14958080053329467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.01,0.12094080448150635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.01,0.4831552028656006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,64,power_law_1.2,0.1679103970527649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.01,0.140339195728302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.01,0.6247744083404541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,64,power_law_1.2,0.22342400550842284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.01,0.1975808024406433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,64,power_law_1.01,0.7831039905548096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,64,power_law_1.2,0.2677504062652588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.01,0.2434623956680298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,64,power_law_1.01,1.2986944198608399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,64,power_law_1.2,0.35727999210357664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.01,0.3555392026901245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,64,power_law_1.01,2.456012725830078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,64,power_law_1.2,0.44471039772033694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.01,0.44307842254638674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,64,power_law_1.2,0.7124224185943604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,64,power_law_1.01,0.6014912128448486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,64,power_law_1.2,0.8768832206726074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,64,power_law_1.01,0.9637887954711915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,64,power_law_1.2,1.0815679550170898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,64,power_law_1.01,1.9766975402832032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,64,power_law_1.2,1.8222272872924805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,64,power_law_1.2,0.0692351996898651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,64,power_law_1.2,3.445804977416992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,64,power_law_1.2,0.0664896011352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,64,power_law_1.2,0.06211199760437012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,64,power_law_1.2,0.06232960224151611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,64,power_law_1.2,0.06695680022239685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,64,power_law_1.2,0.06944000124931335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,64,power_law_1.2,0.07134720087051391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,64,power_law_1.2,0.07219840288162231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,64,power_law_1.2,0.06752640008926392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,64,power_law_1.2,0.07288960218429566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,64,power_law_1.2,0.07483519911766053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,64,power_law_1.2,0.07516160011291503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,64,power_law_1.2,0.07523840069770812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,64,power_law_1.2,0.07575039863586426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,64,power_law_1.2,0.08256000280380249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,64,power_law_1.2,0.08401280045509338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,64,power_law_1.2,0.0913536012172699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,64,power_law_1.2,0.1074112057685852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,64,power_law_1.2,0.11918720006942748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,64,power_law_1.2,0.13817600011825562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,64,power_law_1.2,0.1620736002922058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,64,power_law_1.2,0.22177278995513916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,64,power_law_1.2,0.27393279075622556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,64,power_law_1.2,0.38983678817749023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,64,power_law_1.2,0.4961088180541992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,64,power_law_1.2,0.7438271999359131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,64,power_law_1.2,0.9454591751098633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,64,power_law_1.2,1.2100864410400392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,64,power_law_1.2,1.8954559326171876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,64,power_law_1.2,4.159187316894531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,64,power_law_1.2,0.023187200725078582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,64,power_law_1.2,0.07474560141563416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,64,power_law_1.2,0.023398399353027344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,64,power_law_1.2,0.06812800168991089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,64,power_law_1.2,0.02248319983482361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,64,power_law_1.2,0.05927680134773254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,64,power_law_1.2,0.02332800030708313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,64,power_law_1.2,0.061273598670959474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,64,power_law_1.2,0.0265855997800827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,32,balanced,0.05646933118502299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,64,power_law_1.2,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,32,balanced,0.06065066655476888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,64,power_law_1.2,0.032416000962257385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,32,balanced,0.058821335434913635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,64,power_law_1.2,0.06679040193557739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,64,power_law_1.2,0.05440000295639038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,32,balanced,0.06520000100135803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,64,power_law_1.2,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,64,power_law_1.2,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,32,balanced,0.07764266431331635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,64,power_law_1.2,0.07484800219535828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,64,power_law_1.2,0.05320960283279419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,32,balanced,0.09631466865539551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,32,balanced,0.09359467029571533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,64,power_law_1.2,0.0707584023475647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,64,power_law_1.2,0.052832001447677614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,32,balanced,0.09622400005658467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,64,power_law_1.2,0.07218559980392455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,64,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,32,balanced,0.09473066528638203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,64,power_law_1.2,0.07276800274848938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,64,power_law_1.2,0.052934402227401735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,32,balanced,0.09630933403968811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,64,power_law_1.2,0.07378559708595275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,64,power_law_1.2,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,32,balanced,0.09449600179990132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,64,power_law_1.2,0.07602559924125671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,32,balanced,0.09549867113431294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,64,power_law_1.2,0.051641601324081424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,64,power_law_1.2,0.07404159903526306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,32,balanced,0.0978666643301646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,64,power_law_1.2,0.05308799743652344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,64,power_law_1.2,0.08272640109062195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,32,balanced,0.0937013328075409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,64,power_law_1.2,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,64,power_law_1.2,0.08204159736633301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,32,balanced,0.10129066308339436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,64,power_law_1.2,0.05495679974555969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,64,power_law_1.2,0.08983039855957031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,32,balanced,0.10148266951243083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,64,power_law_1.2,0.052767997980117796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,64,power_law_1.2,0.10315519571304321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,32,balanced,0.10241066416104634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.2,0.0654528021812439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.2,0.11240960359573364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.2,0.08435840010643006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,32,balanced,0.11090667049090068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.2,0.13661439418792726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.2,0.096288001537323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,32,balanced,0.11437867085138957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.2,0.1511296033859253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.2,0.1257472038269043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.2,0.19044480323791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.2,0.15218559503555298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,32,balanced,0.1283466617266337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.2,0.23749759197235107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.2,0.211411190032959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.2,0.31310720443725587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.2,0.2736063957214355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,32,balanced,0.13617600003878275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.2,0.40297598838806153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.2,0.40357117652893065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.2,0.6117311954498291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.2,0.5560768127441407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,32,balanced,0.16054399808247885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.2,0.8367039680480957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,64,power_law_1.2,0.7072832107543945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,64,power_law_1.2,0.9738944053649903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,64,power_law_1.2,1.0653504371643066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,64,power_law_1.2,1.4555520057678222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,32,balanced,0.18124266465504965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,64,power_law_1.2,2.3878976821899416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,64,power_law_1.2,2.994278335571289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,32,balanced,0.052784000833829246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,32,balanced,0.05276266733805338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,32,balanced,0.22807466983795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,32,balanced,0.052832002441088356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,32,balanced,0.05749333401521047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,32,balanced,0.06322666505972545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,32,balanced,0.07205866773923238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,32,balanced,0.07165333131949107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,32,balanced,0.07330133517583211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,32,balanced,0.26926400264104206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,32,balanced,0.07389333347479503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,32,balanced,0.07327466706434886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,32,balanced,0.07801066835721333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,32,balanced,0.07615466912587483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,32,balanced,0.0780320018529892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,32,balanced,0.07950399816036224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,32,balanced,0.0867146650950114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,32,balanced,0.08743466933568318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,32,balanced,0.3770773410797119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,32,balanced,0.09377066294352214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,32,balanced,0.10154133041699727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,32,balanced,0.11126400033632915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,32,balanced,0.13251200318336487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,32,balanced,0.1513706644376119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,32,balanced,0.4666026830673218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,32,balanced,0.1872373421986898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,32,balanced,0.22322134176890054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,32,balanced,0.300709327061971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,32,balanced,0.5756533145904541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,32,balanced,0.3698720137278239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,32,balanced,0.5288373231887817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,32,balanced,0.8558932940165201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,32,balanced,0.6723840236663818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,32,balanced,0.8267839749654134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,32,balanced,0.05217066903909048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,32,balanced,1.275264024734497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,32,balanced,0.05515733361244202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,32,balanced,0.05355200171470642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,32,balanced,1.6570080121358235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,32,balanced,0.059445331494013466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,32,balanced,0.06568000217278798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,32,balanced,0.02325333406527837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,32,balanced,0.023765332996845245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,32,balanced,0.02489600082238515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,32,balanced,0.024383999407291412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,32,balanced,0.024773334463437397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,32,balanced,0.02604266752799352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,32,balanced,0.03397866586844126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,32,balanced,0.03289066751797994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,32,balanced,0.05706666906674703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,32,balanced,0.05669333537419637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,32,balanced,0.05694933235645294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,32,balanced,0.05502399802207947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,32,balanced,0.05316799879074097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,32,balanced,0.051402668158213295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,32,balanced,0.06169066826502482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,32,balanced,0.06237866481145223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,32,balanced,0.06190933287143707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,32,balanced,0.07170133292675018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,32,balanced,0.081386665503184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,32,balanced,0.10110400120417277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,32,balanced,0.1165066659450531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,32,balanced,0.1518346667289734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,32,balanced,0.18863467375437418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,32,balanced,2.4977919260660806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,32,balanced,0.2566666603088379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,32,balanced,0.33211733897527057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,32,balanced,0.4828266700108846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,32,balanced,0.6453866561253866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,32,balanced,0.8051733175913492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,32,power_law_1.01,0.09285759925842285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,32,power_law_1.01,0.08902400135993957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,32,power_law_1.01,0.07896320223808288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,32,power_law_1.01,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,32,power_law_1.01,0.07917439937591553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,32,power_law_1.01,0.08187519907951354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,32,power_law_1.01,0.0895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,32,power_law_1.01,0.09093120098114013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,32,power_law_1.01,0.08972799777984619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,32,power_law_1.01,0.09223679900169372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,32,power_law_1.01,0.09355520009994507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,32,power_law_1.01,0.09285119771957398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,32,balanced,1.2588853041330974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,32,power_law_1.01,0.09550079703330994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,32,power_law_1.01,0.09463040232658386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,32,power_law_1.01,0.09982720017433167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,32,power_law_1.01,0.09850879907608032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,32,power_law_1.01,0.10275839567184449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,32,power_law_1.01,0.11160320043563843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,32,power_law_1.01,0.12254079580307006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,32,power_law_1.01,0.13665920495986938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,32,balanced,0.07789866626262665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,32,power_law_1.01,0.15043840408325196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,32,balanced,0.07558933397134145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,32,balanced,0.07712533573309581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,32,power_law_1.01,0.19356160163879393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,32,balanced,0.075914666056633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,32,power_law_1.01,0.2330944061279297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,32,balanced,0.07551999886830647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,32,power_law_1.01,0.30681600570678713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,32,balanced,0.07784000039100647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,32,balanced,0.07749333480993907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,32,power_law_1.01,0.37868800163269045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,32,balanced,0.07919999957084656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,32,power_law_1.01,0.5299903869628906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,32,balanced,0.07948266466458638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,32,power_law_1.01,0.620959997177124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,32,balanced,0.08287466565767924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,32,power_law_1.01,0.7538559913635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,32,balanced,0.08335999647776286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,32,power_law_1.01,1.2343104362487793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,32,balanced,0.08795733253161113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,32,power_law_1.01,0.07488639950752259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,32,power_law_1.01,0.06844800114631652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,32,balanced,0.09227200349171956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,32,power_law_1.01,2.256166458129883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,32,power_law_1.01,0.06555520296096802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,32,balanced,0.09829866886138916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,32,power_law_1.01,0.05935360193252563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,32,power_law_1.01,0.06474239826202392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,32,balanced,0.11537599563598633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,32,power_law_1.01,0.06589440107345582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,32,power_law_1.01,0.07041280269622803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,32,balanced,0.1292586624622345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,32,power_law_1.01,0.07155839800834655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,32,power_law_1.01,0.07297279834747314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,32,power_law_1.01,0.07043200135231018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,32,balanced,0.1564533313115438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,32,power_law_1.01,0.07370240092277527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,32,power_law_1.01,0.07581440210342408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,32,power_law_1.01,0.07567359805107117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,32,balanced,0.18201067050298056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,32,power_law_1.01,0.07580800056457519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,32,power_law_1.01,0.08257920145988465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,32,power_law_1.01,0.0855679988861084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,32,power_law_1.01,0.09073920249938965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,32,power_law_1.01,0.1059391975402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,32,balanced,0.24261866013209024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,32,power_law_1.01,0.11545599699020385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,32,power_law_1.01,0.13978879451751708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,32,power_law_1.01,0.16149120330810546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,32,power_law_1.01,0.2099008083343506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,32,balanced,2.4825226465861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,32,power_law_1.01,0.262444806098938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,32,balanced,0.29147199789683026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,32,power_law_1.01,0.34452478885650634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,32,power_law_1.01,0.4484799861907959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,32,power_law_1.01,0.6450560092926025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,32,power_law_1.01,0.816864013671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,32,power_law_1.01,1.015174388885498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,32,power_law_1.01,1.5694463729858399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,32,balanced,0.4040373166402181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,32,power_law_1.01,3.1052799224853516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,32,balanced,0.5158720016479492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,32,balanced,0.6270453135172526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,32,balanced,0.9568586349487305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,32,power_law_1.01,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,32,power_law_1.01,0.07304319739341736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,32,power_law_1.01,0.06579840183258057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,32,power_law_1.01,0.0602944016456604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,32,power_law_1.01,0.06316800117492676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,32,power_law_1.01,0.06810240149497986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,32,power_law_1.01,0.07404800057411194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,32,power_law_1.01,0.07061759829521179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,32,power_law_1.01,0.07403519749641418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,32,power_law_1.01,0.07315840125083924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,32,power_law_1.01,0.07408639788627625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,32,power_law_1.01,0.07521920204162598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,32,power_law_1.01,0.07896959781646729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,32,power_law_1.01,0.07828480005264282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,32,power_law_1.01,0.08273280262947083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,32,power_law_1.01,0.08431360125541687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,32,power_law_1.01,0.08856319785118102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,32,power_law_1.01,0.0992959976196289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.01,0.11044479608535766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.01,0.1282431960105896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.01,0.14249600172042848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.01,0.18224639892578126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.01,0.21319680213928222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.01,0.28908801078796387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.01,0.3644416093826294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,32,power_law_1.01,0.022419199347496033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.01,0.5156735897064209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,32,power_law_1.2,0.09484800100326538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,32,power_law_1.01,0.024953599274158477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,32,balanced,1.8765920003255208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.01,0.6690944194793701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,32,power_law_1.2,0.09121919870376587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,32,power_law_1.01,0.021715199947357176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,32,power_law_1.01,0.8052543640136719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,32,power_law_1.2,0.06931840181350708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,32,power_law_1.01,0.02423679977655411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,32,power_law_1.2,0.07678719758987426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,32,power_law_1.01,1.261747169494629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,32,power_law_1.01,0.02279680073261261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,32,power_law_1.2,0.08049280047416688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,32,power_law_1.01,0.025804799795150758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,32,power_law_1.01,2.429849624633789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,32,power_law_1.2,0.08448640108108521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,32,power_law_1.01,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,32,power_law_1.2,0.09011840224266052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,32,power_law_1.01,0.03208959996700287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,32,power_law_1.2,0.0899071991443634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,32,power_law_1.01,0.05557119846343994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,32,power_law_1.2,0.09203199744224548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,32,power_law_1.01,0.05483520030975342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,32,power_law_1.2,0.09130240082740784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,32,power_law_1.01,0.05496320128440857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,32,power_law_1.2,0.09503999948501587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,32,power_law_1.01,0.05443840026855469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,32,power_law_1.2,0.09741439819335937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,32,power_law_1.01,0.055232000350952146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,32,power_law_1.2,0.0941760003566742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,32,power_law_1.2,0.09704319834709167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,32,power_law_1.01,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,32,power_law_1.2,0.09793919920921326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,32,power_law_1.01,0.053401601314544675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,32,power_law_1.2,0.10375679731369018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,32,power_law_1.01,0.057606399059295654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,32,power_law_1.2,0.10666879415512084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,32,power_law_1.01,0.05793280005455017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,32,power_law_1.2,0.11510399580001832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,32,power_law_1.01,0.058233600854873654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,32,power_law_1.2,0.12388479709625244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.01,0.061945599317550656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,32,power_law_1.2,0.1420032024383545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.01,0.07345920205116271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,32,power_law_1.2,0.16343679428100585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.01,0.08625919818878174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,32,power_law_1.2,0.2095103979110718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.01,0.11246720552444459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,32,power_law_1.2,0.23653120994567872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.01,0.13000960350036622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,32,power_law_1.2,0.32730879783630373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.01,0.17301119565963746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,32,power_law_1.2,0.43841919898986814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.01,0.22435200214385986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,32,power_law_1.2,0.625267219543457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.01,0.3068288087844849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,32,power_law_1.2,0.8905535697937011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.01,0.4239999771118164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,32,power_law_1.2,1.031123161315918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,32,power_law_1.01,0.5134335994720459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,32,power_law_1.2,1.5627391815185547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,32,power_law_1.01,0.768230390548706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,32,power_law_1.2,3.420787048339844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,32,power_law_1.01,1.5017919540405273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,32,power_law_1.2,0.07544320225715637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,32,power_law_1.2,0.06977279782295227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,32,power_law_1.2,0.05619199872016907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,32,power_law_1.2,0.061388802528381345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,32,power_law_1.2,0.0791104018688202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,32,power_law_1.2,0.06512640118598938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,32,power_law_1.2,0.0722495973110199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,32,power_law_1.2,0.06783360242843628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,32,power_law_1.2,0.05729280114173889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,32,power_law_1.2,0.07003520131111145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,32,power_law_1.2,0.05975679755210876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,32,power_law_1.2,0.07349119782447815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,32,power_law_1.2,0.062489598989486694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,32,power_law_1.2,0.0706816017627716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,32,power_law_1.2,0.06976640224456787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,32,power_law_1.2,0.07269759774208069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,32,power_law_1.2,0.07128959894180298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,32,power_law_1.2,0.07399680018424988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,32,power_law_1.2,0.07296000123023987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,32,power_law_1.2,0.07571840286254883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,32,power_law_1.2,0.07486720085144043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,32,power_law_1.2,0.0753600001335144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,32,power_law_1.2,0.07432320117950439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,32,power_law_1.2,0.0804095983505249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,32,power_law_1.2,0.07315199971199035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,32,power_law_1.2,0.08760960102081299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,32,power_law_1.2,0.07560960054397584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,32,power_law_1.2,0.08684800267219543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,32,power_law_1.2,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,32,power_law_1.2,0.09548799991607666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,32,power_law_1.2,0.07823359966278076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,32,power_law_1.2,0.10723199844360351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,32,power_law_1.2,0.08252800107002259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,32,power_law_1.2,0.12053120136260986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,32,power_law_1.2,0.08386560082435608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,32,power_law_1.2,0.14354560375213624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,32,power_law_1.2,0.08603519797325135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,32,power_law_1.2,0.17596800327301027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,32,power_law_1.2,0.10279040336608887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,32,power_law_1.2,0.23111679553985595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.2,0.11674879789352417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,32,power_law_1.2,0.2700927972793579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.2,0.1355520009994507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,32,power_law_1.2,0.3718656063079834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.2,0.15300480127334595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,32,power_law_1.2,0.4904319763183594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.2,0.19737600088119506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,32,power_law_1.2,0.7273087978363038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.2,0.24474239349365234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,32,power_law_1.2,0.9404159545898437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.2,0.3206592082977295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,32,power_law_1.2,1.2254719734191895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,16,balanced,0.056554665168126426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.2,0.40981760025024416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,32,power_law_1.2,1.8786624908447265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,16,balanced,0.057520002126693726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.2,0.6161983966827392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,16,balanced,0.06122666597366333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,32,power_law_1.2,3.3750080108642577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,16,balanced,0.06684266527493794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.2,0.743123197555542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,32,power_law_1.2,0.02237440049648285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,16,balanced,0.07592533528804779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,32,power_law_1.2,0.9515839576721191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,16,balanced,0.10082667072614034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,32,power_law_1.2,0.024723200500011443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,16,balanced,0.0969493289788564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,32,power_law_1.2,1.548044776916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,32,power_law_1.2,0.022316800057888032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,16,balanced,0.1002400020758311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,16,balanced,0.09717333316802979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,32,power_law_1.2,0.02276480048894882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,32,power_law_1.2,2.9825408935546873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,16,balanced,0.10016533732414246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,32,power_law_1.2,0.021753600239753722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,16,balanced,0.09903466701507568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,32,power_law_1.2,0.02563199996948242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,16,balanced,0.09876799583435059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,32,power_law_1.2,0.03170560002326965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,16,balanced,0.09871466954549153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,32,power_law_1.2,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,16,balanced,0.09986666838328044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,32,power_law_1.2,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,16,balanced,0.10232533017794292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,16,balanced,0.10479999581972758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,32,power_law_1.2,0.05368319749832153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,32,power_law_1.2,0.05475839972496033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,16,balanced,0.10990933577219646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,32,power_law_1.2,0.05416960120201111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,16,balanced,0.11392000317573547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,32,power_law_1.2,0.05322239995002746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,16,balanced,0.12056000034014384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,32,power_law_1.2,0.05312640070915222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,32,power_law_1.2,0.05248000025749207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,16,balanced,0.13436266779899597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,32,power_law_1.2,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,32,power_law_1.2,0.056704002618789676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,16,balanced,0.14627200365066528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,32,power_law_1.2,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.2,0.06325119733810425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.2,0.07585920095443725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,16,balanced,0.17148266235987344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.2,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.2,0.11706240177154541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.2,0.1373311996459961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,16,balanced,0.19170665740966797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.2,0.18670079708099366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.2,0.2375040054321289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.2,0.34824318885803224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.2,0.4395264148712158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,16,balanced,0.25031999746958417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,32,power_law_1.2,0.5716288089752197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,32,power_law_1.2,0.8910911560058594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,32,power_law_1.2,1.8130880355834962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,16,balanced,0.2969653407732646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,16,balanced,0.42002665996551514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,16,balanced,0.5248266855875651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,16,balanced,0.6403040091196696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,16,balanced,0.05031466484069824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,16,balanced,0.053264002005259194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,16,balanced,0.053157334526379905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,16,balanced,0.9677493572235107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,16,balanced,0.05951466659704844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,16,balanced,0.064410666624705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,16,balanced,0.07411733269691467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,16,balanced,0.07984533409277599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,16,balanced,0.07937600215276082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,16,balanced,0.07915733257929485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,16,balanced,0.07937066753705342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,16,balanced,0.08134933312733968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,16,balanced,0.08107733229796092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,16,balanced,0.05022400120894114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,16,balanced,0.05287466446558634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,16,balanced,0.08024533092975616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,16,balanced,0.055205335219701133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,16,balanced,0.08708799878756206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,16,balanced,0.058880001306533813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,16,balanced,0.09310400485992432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,16,balanced,0.06779733300209045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,16,balanced,0.09425600369771321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,16,balanced,0.081386665503184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,16,balanced,0.08088533580303192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,16,balanced,0.09742933511734009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,16,balanced,0.08286933104197185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,16,balanced,0.11219200491905212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,16,balanced,0.08593599994977315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,16,balanced,0.08074666559696198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,16,balanced,0.11949333548545837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,16,balanced,0.08259200056393941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,16,balanced,0.0843946635723114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,16,balanced,0.1416266659895579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,16,balanced,0.08343467116355896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,16,balanced,0.08679466446240743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,16,balanced,0.16039466857910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,16,balanced,0.09098666906356812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,16,balanced,0.09322667121887207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,16,balanced,0.09574400385220845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,16,balanced,0.20028799772262573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,16,balanced,0.10456533233324687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,16,balanced,0.10982400178909302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,16,balanced,0.23817066351572672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,16,balanced,0.1275200049082438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,16,balanced,0.14421332875887552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,16,balanced,0.3272320032119751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,16,balanced,0.17265599966049194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,16,balanced,1.8690667152404785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,16,balanced,0.4007733265558879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,16,balanced,0.19957866271336874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,16,balanced,0.023370665808518726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,16,balanced,0.024720000723997753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,16,balanced,0.02420266717672348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,16,balanced,0.024645333488782246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,16,balanced,0.02613866577545802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,16,balanced,0.2719306747118632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,16,balanced,0.028607999285062153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,16,balanced,0.5757226546605428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,16,balanced,0.031104000906149547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,16,balanced,0.03062933435042699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,16,balanced,0.040362666050593056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,16,balanced,0.040175999204317726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,16,balanced,0.03933866570393244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,16,balanced,0.06098133325576782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,16,balanced,0.06078400214513143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,16,balanced,0.06253333389759064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,16,balanced,0.3250826597213745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,16,balanced,0.06228266656398773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,16,balanced,0.06425066788991292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,16,balanced,0.06782400111357371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,16,balanced,0.734224001566569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,16,balanced,0.07725333174069722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,16,balanced,0.08260266482830048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,16,balanced,0.09648000200589497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,16,balanced,0.11475200454394023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,16,balanced,0.4577866792678833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,16,balanced,0.14378666877746582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,16,balanced,0.9054826895395914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,16,balanced,0.17932800451914468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,16,balanced,0.22495466470718384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,16,balanced,0.5798399845759074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,16,balanced,0.27963199218114215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,16,balanced,1.3969705899556477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,16,balanced,0.4179413318634033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,16,balanced,0.7129866282145182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,16,balanced,0.5331626733144125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,16,balanced,0.6801599661509196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,16,balanced,1.0970826943715413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,16,power_law_1.01,0.08253440260887146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,16,balanced,2.737802823384603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,16,power_law_1.01,0.09959679841995239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,16,power_law_1.01,0.07267199754714966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,16,power_law_1.01,0.07548800110816956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,16,power_law_1.01,0.0815999984741211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,16,balanced,1.0649759769439697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,16,power_law_1.01,0.08508800268173218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,16,power_law_1.01,0.08958719968795777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,16,power_law_1.01,0.09287040233612061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,16,power_law_1.01,0.09093760251998902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,16,power_law_1.01,0.09362559914588928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,16,power_law_1.01,0.09466879963874816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,16,power_law_1.01,0.0954367995262146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,16,power_law_1.01,0.09659519791603088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,16,power_law_1.01,0.09881600141525268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,16,power_law_1.01,0.10294400453567505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,16,power_law_1.01,0.10400639772415161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,16,power_law_1.01,0.10821759700775146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,16,power_law_1.01,0.11759359836578369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,16,power_law_1.01,0.125273597240448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,16,power_law_1.01,0.14874240159988403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,16,power_law_1.01,0.16155519485473632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,16,power_law_1.01,0.20031359195709228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,16,power_law_1.01,0.23445119857788085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,16,power_law_1.01,0.30593280792236327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,16,power_law_1.01,0.3899519920349121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,16,power_law_1.01,0.5523839950561523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,16,power_law_1.01,0.6558847904205323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,16,power_law_1.01,0.8385343551635742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,16,power_law_1.01,1.3588800430297852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,16,balanced,2.1462507247924805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,16,power_law_1.01,2.656716728210449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,16,balanced,2.076453367869059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,16,power_law_1.01,0.06859520077705383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,16,power_law_1.01,0.08213760256767273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,16,power_law_1.01,0.06505600214004517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,16,power_law_1.01,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,16,power_law_1.01,0.06626560091972351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,16,power_law_1.01,0.06852480173110961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,16,power_law_1.01,0.07605760097503662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,16,power_law_1.01,0.07825919985771179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,16,power_law_1.01,0.0796992003917694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,16,power_law_1.01,0.08081279993057251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,16,power_law_1.01,0.08118399977684021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,16,power_law_1.01,0.0787775993347168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,16,power_law_1.01,0.08292480111122132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,16,power_law_1.01,0.08625280261039733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,16,power_law_1.01,0.09113600254058837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,16,power_law_1.01,0.09389439821243287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,16,power_law_1.01,0.10035200119018554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,16,power_law_1.01,0.11269760131835938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,16,power_law_1.01,0.1259392023086548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,16,power_law_1.01,0.14999040365219116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,16,power_law_1.01,0.17294080257415773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,16,power_law_1.01,0.22376320362091065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,16,power_law_1.01,0.2771903991699219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,16,power_law_1.01,0.3780287981033325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,16,power_law_1.01,0.47216000556945803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,16,power_law_1.01,0.670636796951294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,16,power_law_1.01,0.8484607696533203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,16,power_law_1.01,1.070963191986084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,16,power_law_1.01,1.6635520935058594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,16,power_law_1.01,0.0713919997215271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,16,power_law_1.01,3.3047744750976564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,16,power_law_1.01,0.08858240246772767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,16,power_law_1.2,0.07856000065803528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,16,power_law_1.01,0.05898879766464234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,16,power_law_1.2,0.09678720235824585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,16,power_law_1.01,0.0642687976360321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,16,power_law_1.01,0.022265599668025972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,16,power_law_1.2,0.06871039867401123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,16,power_law_1.01,0.06690559983253479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,16,power_law_1.01,0.023552000522613525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,16,power_law_1.2,0.07566720247268677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,16,power_law_1.01,0.0726527988910675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,16,power_law_1.01,0.023628799617290495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,16,power_law_1.2,0.07788159847259521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,16,power_law_1.01,0.07644799947738648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,16,power_law_1.01,0.02369280010461807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,16,power_law_1.2,0.0813696026802063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,16,power_law_1.01,0.07964159846305847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,16,power_law_1.01,0.025183999538421632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,16,power_law_1.01,0.08010240197181702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,16,power_law_1.2,0.08963840007781983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,16,power_law_1.01,0.025484800338745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,16,power_law_1.01,0.07813119888305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,16,power_law_1.2,0.0929535984992981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,16,power_law_1.01,0.026214399933815004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,16,power_law_1.01,0.07944959998130799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,16,power_law_1.2,0.09240319728851318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,16,power_law_1.01,0.027526399493217467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,16,power_law_1.01,0.08258559703826904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,16,power_law_1.2,0.0936896026134491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,16,power_law_1.01,0.03344640135765076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,16,power_law_1.01,0.08330240249633789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,16,power_law_1.2,0.09590399861335755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,16,power_law_1.01,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,16,power_law_1.01,0.08553599715232849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,16,power_law_1.2,0.09719039797782898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,16,power_law_1.01,0.03452160060405731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,16,power_law_1.01,0.0919871985912323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,16,power_law_1.2,0.09573760032653808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,16,power_law_1.01,0.09442560076713562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,16,power_law_1.01,0.057107198238372806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,16,power_law_1.2,0.09720320105552674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,16,power_law_1.01,0.09644160270690919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,16,power_law_1.01,0.05854719877243042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,16,power_law_1.2,0.10500479936599731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,16,power_law_1.01,0.10728319883346557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,16,power_law_1.01,0.0609279990196228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,16,power_law_1.2,0.10540800094604492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.01,0.12136319875717164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,16,power_law_1.01,0.055936002731323244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,16,power_law_1.2,0.11338239908218384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.01,0.14120320081710816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,16,power_law_1.01,0.05935360193252563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,16,power_law_1.2,0.12301440238952636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.01,0.15889279842376708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,16,power_law_1.01,0.06034560203552246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,16,power_law_1.2,0.12823679447174072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.01,0.20360960960388183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,16,power_law_1.01,0.0650111973285675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,16,power_law_1.2,0.15015679597854614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.01,0.24104959964752198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.01,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,16,power_law_1.2,0.16625920534133912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.01,0.3259007930755615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.01,0.07652480006217957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,16,power_law_1.2,0.20762879848480226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.01,0.41841278076171873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.01,0.08499199748039246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,16,power_law_1.2,0.25861120223999023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.01,0.10384639501571655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.01,0.5501887798309326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,16,power_law_1.2,0.3235327959060669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.01,0.13506560325622557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.01,0.7560448169708252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,16,power_law_1.2,0.43099517822265626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.01,0.1723647952079773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,16,power_law_1.01,0.8819456100463867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,16,power_law_1.2,0.6216256141662597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.01,0.2142400026321411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,16,power_law_1.01,1.357363224029541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,16,power_law_1.2,0.7428544044494629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.01,0.30125439167022705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,16,power_law_1.2,0.9914560317993164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,16,power_law_1.01,2.736537551879883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.01,0.3957632064819336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,16,power_law_1.2,1.5712512016296387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,16,power_law_1.01,0.46985602378845215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,16,power_law_1.2,2.9183488845825196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,16,power_law_1.01,0.7395711898803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,16,power_law_1.01,1.409113597869873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,16,power_law_1.2,0.06861439943313599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,16,power_law_1.2,0.07752320170402527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,16,power_law_1.2,0.05682560205459595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,16,power_law_1.2,0.06165760159492493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,16,power_law_1.2,0.06565120220184326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,16,power_law_1.2,0.06774399876594543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,16,power_law_1.2,0.07377279996871948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,16,power_law_1.2,0.07736960053443909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,16,power_law_1.2,0.07702400088310242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,16,power_law_1.2,0.07728639841079712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,16,power_law_1.2,0.08216959834098816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,16,power_law_1.2,0.0824895977973938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,16,power_law_1.2,0.08389760255813598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,16,power_law_1.2,0.08625280261039733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,16,power_law_1.2,0.09251840114593506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,16,power_law_1.2,0.09477760195732117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,16,power_law_1.2,0.10119040012359619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,16,power_law_1.2,0.11464320421218872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,16,power_law_1.2,0.12878079414367677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,16,power_law_1.2,0.1553984045982361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,16,power_law_1.2,0.1768447995185852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,16,power_law_1.2,0.23978879451751708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,16,power_law_1.2,0.2927999973297119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,16,power_law_1.2,0.39663360118865965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,16,power_law_1.2,0.5018239974975586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,8,balanced,0.05446400245030721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,16,power_law_1.2,0.7119872093200683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,16,power_law_1.2,0.02144639939069748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,8,balanced,0.058917333682378135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,8,balanced,0.06135466694831848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,16,power_law_1.2,0.023027199506759643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,16,power_law_1.2,1.0140031814575194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,8,balanced,0.06781333188215892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,16,power_law_1.2,0.023417599499225616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,16,power_law_1.2,1.1913663864135742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,8,balanced,0.0827893316745758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,16,power_law_1.2,0.07070720195770264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,16,power_law_1.2,0.023001599311828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,8,balanced,0.10947199662526448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,16,power_law_1.2,1.875833511352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,8,balanced,0.10814932982126872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,16,power_law_1.2,0.08601599931716919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,16,power_law_1.2,0.025203201174736022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,8,balanced,0.1090186635653178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,16,power_law_1.2,0.05751680135726929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,16,power_law_1.2,0.02459519952535629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,16,power_law_1.2,3.756390380859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,8,balanced,0.11064533392588298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,16,power_law_1.2,0.060192000865936277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,16,power_law_1.2,0.027456000447273254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,8,balanced,0.10498666763305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,16,power_law_1.2,0.06751999855041504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,8,balanced,0.10408533612887065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,16,power_law_1.2,0.027987200021743774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,8,balanced,0.1067733367284139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,16,power_law_1.2,0.06963199973106385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,16,power_law_1.2,0.03540480136871338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,8,balanced,0.10539199908574422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,16,power_law_1.2,0.07648640275001525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,16,power_law_1.2,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,8,balanced,0.10846933722496033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,16,power_law_1.2,0.08042880296707153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,16,power_law_1.2,0.035180801153182985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,8,balanced,0.11437867085138957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,16,power_law_1.2,0.07874559760093688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,16,power_law_1.2,0.05625600218772888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,8,balanced,0.11389333009719849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,16,power_law_1.2,0.07996799945831298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,16,power_law_1.2,0.05921279788017273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,8,balanced,0.11885333061218262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,16,power_law_1.2,0.08048639893531799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,16,power_law_1.2,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,8,balanced,0.12942933042844137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,16,power_law_1.2,0.08242560029029847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,16,power_law_1.2,0.05717759728431702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,8,balanced,0.13461333513259888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,16,power_law_1.2,0.08205440044403076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,16,power_law_1.2,0.05898879766464234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,16,power_law_1.2,0.08421120047569275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,16,power_law_1.2,0.06284800171852112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,8,balanced,0.15265599886576334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,16,power_law_1.2,0.09029120206832886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,16,power_law_1.2,0.06428800225257873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,16,power_law_1.2,0.09412479996681214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.2,0.06850559711456299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,8,balanced,0.16729599237442017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,16,power_law_1.2,0.1013759970664978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.2,0.0780672013759613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,16,power_law_1.2,0.11384320259094238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.2,0.0897599995136261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.2,0.1258944034576416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,8,balanced,0.19555733601252237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.2,0.11443840265274048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.2,0.14274560213088988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.2,0.14028799533843994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.2,0.16408319473266603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.2,0.18805760145187378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.2,0.20541439056396485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.2,0.22766079902648925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,8,balanced,0.22370133797327676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.2,0.24245119094848633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.2,0.32753920555114746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.2,0.35997440814971926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.2,0.4256256103515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.2,0.4372543811798096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,16,power_law_1.2,0.521011209487915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,8,balanced,0.3007146716117859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.2,0.6363584041595459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,16,power_law_1.2,0.8023232460021973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.2,0.8191807746887207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,16,power_law_1.2,1.532857608795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,16,power_law_1.2,1.0112704277038573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,16,power_law_1.2,1.6200832366943358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,8,balanced,0.3617386817932129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,16,power_law_1.2,3.176134490966797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,8,balanced,0.5176639954249064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,8,balanced,0.6484373410542806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,8,balanced,0.8003199895222982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,8,balanced,1.204026699066162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,8,balanced,0.05048533280690511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,8,balanced,0.05495999753475189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,8,balanced,0.057445332407951355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,8,balanced,0.06305066744486491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,8,balanced,0.068271999557813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,8,balanced,0.08948799967765808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,8,balanced,0.08388800422350566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,8,balanced,0.08370666702588399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,8,balanced,0.08629866441090901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,8,balanced,0.08832533160845439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,8,balanced,0.08746133248011272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,8,balanced,0.0890880028406779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,8,balanced,0.09295466542243958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,8,balanced,0.0936853289604187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,8,balanced,0.10238933563232422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,8,balanced,0.10548800230026245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,8,balanced,0.11101333300272624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,8,balanced,0.12404800454775493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,8,balanced,0.13571199774742126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,8,balanced,0.1605226695537567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,8,balanced,0.18086934089660645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,8,balanced,0.05256533126036326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,8,balanced,2.335066636403402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,8,balanced,0.22473067045211792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,8,balanced,0.05685333410898844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,8,balanced,0.059621334075927734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,8,balanced,0.06906666855017345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,8,balanced,0.0773119976123174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,8,balanced,0.023365333676338196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,8,balanced,0.2678346633911133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,8,balanced,0.10723732908566792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,8,balanced,0.025253333151340485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,8,balanced,0.10934933026631673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,8,balanced,0.02672533442576726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,8,balanced,0.10678933064142863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,8,balanced,0.028560000161329906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,8,balanced,0.10663466652234395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,8,balanced,0.028805332879225414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,8,balanced,0.10670933127403259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,8,balanced,0.03161599983771642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,8,balanced,0.37488532066345215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,8,balanced,0.03181866556406021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,8,balanced,0.10125333070755005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,8,balanced,0.032816000282764435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,8,balanced,0.1088053286075592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,8,balanced,0.03461333364248276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,8,balanced,0.11202133695284526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,8,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,8,balanced,0.10878933469454448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,8,balanced,0.036346666514873505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,8,balanced,0.11244266231854756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,8,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,8,balanced,0.11247467001279195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,8,balanced,0.047055999437967934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,8,balanced,0.46138131618499756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,8,balanced,0.1186293363571167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,8,balanced,0.05083733300367991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,8,balanced,0.07300800085067749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,8,balanced,0.12891200184822083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,8,balanced,0.07434133191903432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,8,balanced,0.13831999897956848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,8,balanced,0.07715733349323273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,8,balanced,0.1553653379281362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,8,balanced,0.08597866694132487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,8,balanced,0.09314133723576863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,8,balanced,0.17080533504486084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,8,balanced,0.6664533217748007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,8,balanced,0.10688533385594685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,8,balanced,0.21007466316223145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,8,balanced,0.11841600139935811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,8,balanced,0.24133867025375366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,8,balanced,0.14949333667755127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,8,balanced,0.33883734544118244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,8,balanced,0.176639993985494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,8,balanced,0.8550346692403158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,8,balanced,0.40984535217285156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,8,balanced,0.22481600443522134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,8,balanced,0.28457067410151166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,8,balanced,0.5833813349405924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,8,balanced,1.0557386875152588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,8,balanced,0.41465600331624347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,8,balanced,0.7506720225016276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,8,balanced,0.5208213329315186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,8,balanced,1.630309263865153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,8,balanced,0.9189066886901855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,8,balanced,0.6543466647466024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,8,balanced,1.4167466163635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,8,power_law_1.01,0.07194240093231201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,8,power_law_1.01,0.08298239707946778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,8,power_law_1.01,0.07245439887046815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,8,power_law_1.01,0.07502080202102661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,8,power_law_1.01,0.0796992003917694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,8,power_law_1.01,0.0896127998828888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,8,power_law_1.01,0.0952127993106842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,8,power_law_1.01,0.10015360116958619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,8,power_law_1.01,0.09908480048179627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,8,power_law_1.01,0.09947519898414611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,8,power_law_1.01,0.10179200172424316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,8,power_law_1.01,0.10325119495391846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,8,power_law_1.01,0.10252159833908081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,8,power_law_1.01,0.10644479990005493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,8,balanced,1.0187040170033772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,8,power_law_1.01,0.11169919967651368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,8,power_law_1.01,0.11449600458145141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,8,power_law_1.01,0.11843839883804322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,8,power_law_1.01,0.12888959646224976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,8,power_law_1.01,0.14142080545425414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,8,balanced,3.1995681126912436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,8,power_law_1.01,0.16335359811782837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,8,power_law_1.01,0.17843199968338014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,8,power_law_1.01,0.22167038917541504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,8,power_law_1.01,0.26540160179138184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,8,power_law_1.01,0.3574336051940918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,8,power_law_1.01,0.43893117904663087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,8,power_law_1.01,0.6278592109680176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,8,power_law_1.01,0.859347152709961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,8,power_law_1.01,0.9742143630981446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,8,balanced,2.7758560180664062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,8,power_law_1.01,1.458745574951172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,8,power_law_1.01,2.9875200271606444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,8,balanced,2.0082133611043296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,8,power_law_1.01,0.06583679914474487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,8,power_law_1.01,0.07183359861373902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,8,power_law_1.01,0.0580672025680542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,8,power_law_1.01,0.0616703987121582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,8,power_law_1.01,0.06874880194664001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,8,power_law_1.01,0.07839360237121581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,8,power_law_1.01,0.08177279829978942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,8,power_law_1.01,0.07144320011138916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,8,power_law_1.01,0.08320639729499817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,8,power_law_1.01,0.08316799998283386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,8,power_law_1.01,0.0838271975517273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,8,power_law_1.01,0.0215488001704216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,8,power_law_1.01,0.08455680012702942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,8,power_law_1.01,0.06753280162811279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,8,power_law_1.01,0.023423999547958374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,8,power_law_1.01,0.06679040193557739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,8,power_law_1.01,0.08860160112380981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,8,power_law_1.01,0.024857600033283234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,8,power_law_1.01,0.0896448016166687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,8,power_law_1.01,0.07541120052337646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,8,power_law_1.01,0.02508159875869751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,8,power_law_1.01,0.09012479782104492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,8,power_law_1.01,0.08580480217933655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,8,power_law_1.01,0.026393601298332216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,8,power_law_1.01,0.09477120041847228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,8,power_law_1.01,0.09584640264511109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,8,power_law_1.01,0.0293503999710083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,8,power_law_1.01,0.09711999893188476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,8,power_law_1.01,0.10151679515838623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,8,power_law_1.01,0.02922239899635315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,8,power_law_1.01,0.09731199741363525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,8,power_law_1.01,0.10188159942626954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,8,power_law_1.01,0.029145601391792297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,8,power_law_1.01,0.09889919757843017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,8,power_law_1.01,0.11100800037384033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,8,power_law_1.01,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,8,power_law_1.01,0.10611200332641602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,8,power_law_1.01,0.12562559843063353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,8,power_law_1.01,0.03154560029506683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,8,power_law_1.01,0.10305279493331909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,8,power_law_1.01,0.13928960561752318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,8,power_law_1.01,0.03131519854068756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,8,power_law_1.01,0.10215679407119752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,8,power_law_1.01,0.17319040298461913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,8,power_law_1.01,0.03982079923152924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,8,power_law_1.01,0.10759680271148682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,8,power_law_1.01,0.19585280418395995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,8,power_law_1.01,0.04017919898033142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,8,power_law_1.01,0.11317119598388672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,8,power_law_1.01,0.04074240028858185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,8,power_law_1.01,0.25546879768371583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,8,power_law_1.01,0.11393280029296875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,8,power_law_1.01,0.06464639902114869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,8,power_law_1.01,0.30984959602355955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,8,power_law_1.01,0.11626880168914795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,8,power_law_1.01,0.06729599833488464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,8,power_law_1.01,0.4184895992279053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,8,power_law_1.01,0.1313599944114685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,8,power_law_1.01,0.07132160067558288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,8,power_law_1.01,0.5287807941436767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.01,0.14251519441604615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,8,power_law_1.01,0.07331200242042542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,8,power_law_1.01,0.7535935878753662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.01,0.1688447952270508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.01,0.07862399816513062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,8,power_law_1.01,1.037932777404785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.01,0.19332480430603027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.01,0.08628479838371277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,8,power_law_1.01,1.2509504318237306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,8,power_law_1.2,0.07219840288162231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.01,0.24149758815765382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.01,0.09454079866409301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,8,power_law_1.01,1.9707071304321289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,8,power_law_1.2,0.08604159951210022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.01,0.3018687963485718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.01,0.11636480093002319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,8,power_law_1.2,0.06820480227470398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.01,0.39479680061340333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.01,0.1319167971611023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,8,power_law_1.01,3.700057601928711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,8,power_law_1.2,0.07802879810333252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.01,0.4880512237548828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.01,0.1858240008354187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,8,power_law_1.2,0.07958400249481201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.01,0.7061056137084961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.01,0.22556159496307374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,8,power_law_1.2,0.08326399922370911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.01,0.8828351974487305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.01,0.3019200086593628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,8,power_law_1.2,0.09515519738197327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,8,power_law_1.01,1.0872256278991699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.01,0.3959680080413818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,8,power_law_1.2,0.09786239862442017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,8,power_law_1.01,0.48857598304748534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,8,power_law_1.2,0.10113919973373413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,8,power_law_1.01,0.7789567947387696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,8,power_law_1.2,0.0979200005531311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,8,power_law_1.2,0.10147199630737305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,8,power_law_1.01,1.508236789703369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,8,power_law_1.2,0.105132794380188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,8,power_law_1.2,0.1064255952835083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,8,power_law_1.2,0.10836479663848878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,8,power_law_1.2,0.11236480474472046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,8,power_law_1.2,0.11555839776992798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,8,power_law_1.2,0.12156800031661988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,8,power_law_1.2,0.1332479953765869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,8,power_law_1.2,0.14565119743347169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,8,power_law_1.2,0.17087359428405763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,8,power_law_1.2,0.18554240465164185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,8,power_law_1.2,0.23615360260009766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,8,power_law_1.2,0.2721280097961426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,8,power_law_1.2,0.40814719200134275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,8,power_law_1.2,0.46816000938415525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,8,power_law_1.2,0.6550015926361084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,8,power_law_1.2,0.877023983001709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,8,power_law_1.2,1.186451244354248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,8,power_law_1.2,1.6883199691772461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,8,power_law_1.2,3.4548480987548826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,8,power_law_1.2,0.06395519971847534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,8,power_law_1.2,0.07226880192756653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,8,power_law_1.2,0.05591040253639221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,8,power_law_1.2,0.07103360295295716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,8,power_law_1.2,0.06247040033340454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,8,power_law_1.2,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,8,power_law_1.2,0.06847360134124755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,8,power_law_1.2,0.05862399935722351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,8,power_law_1.2,0.07243520021438599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,8,power_law_1.2,0.06599040031433105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,8,power_law_1.2,0.08210560083389282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,8,power_law_1.2,0.07209600210189819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,8,power_law_1.2,0.08321920037269592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,8,power_law_1.2,0.08106880187988282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,8,power_law_1.2,0.08465920090675354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,8,power_law_1.2,0.09325439929962158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,8,power_law_1.2,0.08468480110168457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,8,power_law_1.2,0.095551997423172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,8,power_law_1.2,0.08709120154380798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,8,power_law_1.2,0.09596800208091735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,8,power_law_1.2,0.08746880292892456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,8,power_law_1.2,0.09463679790496826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,8,power_law_1.2,0.0923520028591156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,8,power_law_1.2,0.09781759977340698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,8,power_law_1.2,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,8,power_law_1.2,0.0976639986038208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,8,power_law_1.2,0.10154240131378174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,8,power_law_1.2,0.10349440574645996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,8,power_law_1.2,0.10517120361328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,8,power_law_1.2,0.10291839838027954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,8,power_law_1.2,0.11247999668121338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,8,power_law_1.2,0.10937600135803223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,8,power_law_1.2,0.1273344039916992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,8,power_law_1.01,1.7239360809326172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,8,power_law_1.2,0.11297919750213622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,8,power_law_1.2,0.14239360094070436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,8,power_law_1.01,3.4456192016601563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,8,power_law_1.2,0.12298879623413086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,8,power_law_1.2,0.17647360563278197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,8,power_law_1.2,0.1342463970184326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,8,power_law_1.2,0.20490880012512208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.2,0.14936959743499756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,8,power_law_1.2,0.25569920539855956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.2,0.17694720029830932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,8,power_law_1.2,0.021427200734615327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,8,power_law_1.2,0.33728001117706297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.2,0.1982975959777832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,8,power_law_1.2,0.022809599339962006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,8,power_law_1.2,0.4369664192199707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,8,power_law_1.2,0.02388480007648468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.2,0.2529088020324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,8,power_law_1.2,0.5637119770050049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,8,power_law_1.2,0.0253248006105423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.2,0.3123456001281738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,8,power_law_1.2,0.8795904159545899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,8,power_law_1.2,0.02710399925708771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.2,0.42749438285827634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,8,power_law_1.2,1.0448896408081054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,8,power_law_1.2,0.029529601335525513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.2,0.522707176208496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,8,power_law_1.2,0.027321600914001466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.2,0.7573184013366699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,8,power_law_1.2,1.2983103752136231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,8,power_law_1.2,0.029216000437736513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.2,1.071110439300537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,8,power_law_1.2,2.202284812927246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,8,power_law_1.2,0.03054719865322113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,8,power_law_1.2,1.2592703819274902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,8,power_law_1.2,4.218975830078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,8,power_law_1.2,0.030681601166725157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,8,power_law_1.2,1.9423168182373047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,8,power_law_1.2,0.032576000690460204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,8,power_law_1.2,3.8341312408447266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,8,power_law_1.2,0.03944959938526153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,8,power_law_1.2,0.0392767995595932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,8,power_law_1.2,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,8,power_law_1.2,0.06609280109405517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,8,power_law_1.2,0.06722559928894042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,8,power_law_1.2,0.07171840071678162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,8,power_law_1.2,0.07335039973258972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.2,0.07992960214614868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.2,0.09011200070381165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.2,0.09548799991607666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.2,0.12227840423583984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.2,0.14146560430526733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.2,0.19611519575119019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.2,0.2356544017791748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.2,0.32967679500579833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.2,0.42479357719421384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,8,power_law_1.2,0.5434239864349365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,8,power_law_1.2,0.8087807655334472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,8,power_law_1.2,1.616262435913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,4,balanced,0.06051200131575266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,4,balanced,0.06383466720581055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,4,balanced,0.06564266482988994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,4,balanced,0.07398400207360585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,4,balanced,0.09386133154233296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,4,balanced,0.12168000141779582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,4,balanced,0.12175466616948445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,4,balanced,0.12056533495585124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,4,balanced,0.11920000116030376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,4,balanced,0.12140267093976338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,4,balanced,0.12170132994651794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,4,balanced,0.12299199899037679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,4,balanced,0.12481066584587097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,4,balanced,0.12743467092514038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,4,balanced,0.1311360001564026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,4,balanced,0.13619732856750488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,4,balanced,0.14086932937304178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,4,balanced,0.15337066849072775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,4,balanced,0.053717335065205894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,4,power_law_1.01,0.0700160026550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,4,balanced,0.05758399764696757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,4,balanced,0.16387200355529785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,4,balanced,0.05958933134873708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,4,power_law_1.01,0.07484800219535828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,4,balanced,0.07172266642252605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,4,power_law_1.01,0.07253119945526124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,4,balanced,0.08179733157157898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,4,balanced,0.18382932742436728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,4,balanced,0.1092693308989207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,4,power_law_1.01,0.08057600259780884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,4,balanced,0.10843732953071594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,4,power_law_1.01,0.09095680117607116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,4,balanced,0.2018079956372579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,4,balanced,0.1107360025246938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,4,power_law_1.01,0.09754239916801452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,4,balanced,0.11014933387438457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,4,balanced,0.11063466469446818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,4,power_law_1.01,0.10956159830093384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,4,balanced,0.11058132847150166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,4,balanced,0.24395734071731567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,4,power_law_1.01,0.1106943964958191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,4,balanced,0.11552000045776367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,4,power_law_1.01,0.1162943959236145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,4,balanced,0.11764267086982727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,4,power_law_1.01,0.11376639604568481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,4,balanced,0.11868799726168315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,4,power_law_1.01,0.11723519563674926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,4,balanced,0.2863946755727132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,4,balanced,0.1264959971110026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,4,power_law_1.01,0.1177791953086853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,4,balanced,0.12965866923332214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,4,power_law_1.01,0.12007679939270019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,4,balanced,0.13550399740537009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,4,power_law_1.01,0.12164479494094849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,4,balanced,0.1541866660118103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,4,balanced,0.40801600615183514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,4,power_law_1.01,0.12915199995040894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,4,power_law_1.01,0.13222399950027466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,4,balanced,0.16730666160583496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,4,power_law_1.01,0.13644800186157227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,4,balanced,0.19589867194493613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,4,power_law_1.01,0.15642240047454833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,4,power_law_1.01,0.16587519645690918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,4,balanced,0.49262932936350506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,4,power_law_1.01,0.19831039905548095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,4,balanced,0.22130133708318075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,4,power_law_1.01,0.21945600509643554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,4,power_law_1.01,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,4,power_law_1.01,0.2725888013839722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,4,power_law_1.01,0.07036799788475037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,4,power_law_1.01,0.34337921142578126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,4,balanced,0.27915199597676593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,4,power_law_1.01,0.060601598024368285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,4,power_law_1.01,0.45027837753295896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,4,power_law_1.01,0.06636160016059875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,4,power_law_1.01,0.6111423969268799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,4,power_law_1.01,0.08010879755020142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,4,power_law_1.01,0.0866815984249115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,4,power_law_1.01,0.8189696311950684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,4,balanced,0.7188053131103516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,4,balanced,0.3320639928181966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,4,power_law_1.01,0.09870719909667969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,4,power_law_1.01,1.074732780456543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,4,power_law_1.01,0.10440959930419921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,4,power_law_1.01,1.2712320327758788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,4,power_law_1.01,0.10057599544525146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,4,power_law_1.01,1.925811195373535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,4,power_law_1.01,0.1021888017654419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,4,power_law_1.01,0.10412800312042236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,4,power_law_1.01,3.8333248138427733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,4,balanced,0.48505600293477374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,4,power_law_1.01,0.11101440191268921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,4,power_law_1.01,0.11127680540084839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,4,power_law_1.01,0.11645439863204957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,4,balanced,0.9086026350657145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,4,power_law_1.01,0.11896320581436157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,4,power_law_1.01,0.12753920555114745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,4,power_law_1.01,0.13445119857788085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,4,balanced,0.6001493136088053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,4,power_law_1.01,0.1536064028739929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,4,power_law_1.01,0.17621760368347167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,4,power_law_1.01,0.20663039684295653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,4,power_law_1.01,0.24051198959350586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,4,power_law_1.01,0.31089279651641843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,4,power_law_1.01,0.37260799407958983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,4,power_law_1.01,0.5134335994720459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,4,balanced,1.1130506992340088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,4,power_law_1.01,0.6894591808319092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,4,power_law_1.01,0.9534079551696777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,4,balanced,0.8685546716054281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,4,power_law_1.01,1.2522303581237793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,4,power_law_1.01,1.5325440406799316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,4,power_law_1.01,2.5083520889282225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,4,power_law_1.01,4.830707168579101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,4,balanced,1.1303839683532715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,4,balanced,1.7134613990783691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,4,balanced,1.3931999206542969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,4,balanced,3.334773381551107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,4,balanced,2.1571680704752603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,4,power_law_1.2,0.0692799985408783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,4,power_law_1.2,0.07464320063591004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,4,power_law_1.2,0.06659839749336242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,4,power_law_1.2,0.0794368028640747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,4,power_law_1.2,0.089273601770401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,4,power_law_1.2,0.09452160000801087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,4,power_law_1.2,0.1076159954071045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,4,power_law_1.2,0.11289600133895875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,4,power_law_1.2,0.11242239475250244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,4,power_law_1.2,0.11068160533905029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,4,power_law_1.2,0.11586560010910034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,4,power_law_1.2,0.11911040544509888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,4,power_law_1.2,0.11953920125961304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,4,power_law_1.2,0.12547199726104735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,4,power_law_1.2,0.13108479976654053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,4,power_law_1.2,0.1338752031326294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,4,power_law_1.2,0.14117759466171265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,4,power_law_1.2,0.16020480394363404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,4,power_law_1.2,0.1719231963157654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,4,power_law_1.2,0.204256010055542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,4,power_law_1.2,0.23406078815460205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,4,power_law_1.2,0.06369919776916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,4,power_law_1.2,0.28584320545196534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,4,power_law_1.2,0.07037439942359924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,4,power_law_1.2,0.3554752111434937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,4,power_law_1.2,0.058355200290679934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,4,power_law_1.2,0.4809919834136963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,4,power_law_1.2,0.06572800278663635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,4,power_law_1.2,0.5966464042663574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,4,power_law_1.2,0.07349759936332703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,4,power_law_1.2,0.08384000062942505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,4,power_law_1.2,0.8712320327758789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,4,power_law_1.2,0.09760000109672547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,4,power_law_1.2,1.1328767776489257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,4,power_law_1.2,0.10497920513153076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,4,power_law_1.2,1.4101056098937987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,4,balanced,4.260122617085774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,4,power_law_1.2,0.10566400289535523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,4,power_law_1.2,2.269715118408203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,4,power_law_1.2,0.10133119821548461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,4,power_law_1.2,0.10109440088272095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,4,power_law_1.2,3.923104095458984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,4,power_law_1.2,0.11059839725494384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,4,power_law_1.2,0.11062400341033936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,4,power_law_1.2,0.11724159717559815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,4,power_law_1.2,0.12469760179519654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,4,power_law_1.2,0.12944639921188356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,4,power_law_1.2,0.1373247981071472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,4,power_law_1.2,0.16095999479293824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,4,power_law_1.2,0.17858560085296632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,4,power_law_1.2,0.21342079639434813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,4,power_law_1.2,0.2517888069152832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,4,power_law_1.2,0.3279680013656616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,4,power_law_1.2,0.3944704055786133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,4,power_law_1.2,0.5653247833251953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,4,power_law_1.2,0.7643136024475098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,4,power_law_1.2,1.0777536392211915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,4,power_law_1.2,1.3667903900146485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,4,power_law_1.2,1.6384832382202148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,4,power_law_1.2,2.5366912841796876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,4,power_law_1.2,5.147948837280273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,128,balanced,0.034234667817751564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,128,balanced,0.034501334031422935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,128,balanced,0.03461866577466329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,128,balanced,0.03661333272854487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,128,balanced,0.036746665835380554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,128,balanced,0.038191998998324074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,128,balanced,0.03812800099452337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,128,balanced,0.036474667489528656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,128,balanced,0.03898133337497711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,128,balanced,0.038586666186650596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,128,balanced,0.03824000060558319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,128,balanced,0.03835733234882355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,128,balanced,0.03821333249409994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,128,balanced,0.042170668641726174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,128,balanced,0.04275199770927429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,128,balanced,0.04033066580692927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,128,balanced,0.04224533339341482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,128,balanced,0.04248000184694926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,128,balanced,0.04417066772778829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,128,balanced,0.04450666904449463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,128,balanced,0.046021332343419395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,128,balanced,0.05075199902057648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,128,balanced,0.052629331747690834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,128,balanced,0.05834133426348368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,128,balanced,0.06241066753864288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,128,balanced,0.07977066437403361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,128,balanced,0.03172266731659571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,128,balanced,0.032885332902272545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,128,balanced,0.03389866650104523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,128,balanced,0.03329599897066752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,128,balanced,0.03453333427508672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,128,balanced,0.03675200045108795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,128,balanced,0.03465600063403448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,128,balanced,0.03496533383925756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,128,balanced,0.036144000788529716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,128,balanced,0.03698666642109553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,128,balanced,0.03789866715669632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,128,balanced,0.03844266633192698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,128,balanced,0.03712533414363861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,128,balanced,0.03687999894221624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,128,balanced,0.09354133407274882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,128,balanced,0.038149334490299225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,128,balanced,0.03733866661787033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,128,balanced,0.03849066545565923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,128,balanced,0.036981334288915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,128,balanced,0.03900266687075297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,128,balanced,0.03867200016975403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,128,balanced,0.038575999438762665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,128,balanced,0.04064533362785975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,128,balanced,0.03867200016975403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,128,balanced,0.038362666964530945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,128,balanced,0.03836799909671148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,128,balanced,0.040906667709350586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,128,balanced,0.04057066639264425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,128,balanced,0.04261333247025808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,128,balanced,0.04027733455101649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,128,balanced,0.04264533519744873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,128,balanced,0.04116799930731455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,128,balanced,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,128,balanced,0.042437334855397545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,128,balanced,0.044453332821528115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,128,balanced,0.11178666353225708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,128,balanced,0.04310933252175649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,128,balanced,0.0468746672074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,128,balanced,0.04474133253097534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,128,balanced,0.048309331138928734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,128,balanced,0.04412800073623657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,128,balanced,0.0487306664387385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,128,balanced,0.05247466762860616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,128,balanced,0.050293331344922386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,128,balanced,0.056688000758488975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,128,balanced,0.05468800167242686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,128,balanced,0.06317866841952006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,128,balanced,0.056602666775385536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,128,balanced,0.15338666240374246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,128,balanced,0.07550933460394542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,128,balanced,0.0710453341404597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,128,balanced,0.01942933350801468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,128,balanced,0.02091199904680252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,128,balanced,0.020703999946514767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,128,balanced,0.022570667167504627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,128,balanced,0.03295466552178065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,128,balanced,0.030821333328882854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,128,balanced,0.029546665648619335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,128,balanced,0.07994133234024048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,128,balanced,0.02812800059715907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,128,balanced,0.030250666042168934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,128,balanced,0.09628267089525859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,128,balanced,0.03017599880695343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,128,balanced,0.02937600016593933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,128,balanced,0.029722665747006733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,128,balanced,0.03031466652949651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,128,balanced,0.03086400032043457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,128,balanced,0.031914666295051575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,128,balanced,0.03259733319282532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,128,balanced,0.10513066252072652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,128,balanced,0.03256533294916153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,128,balanced,0.03421333432197571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,128,balanced,0.1349546710650126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,128,balanced,0.0367253323396047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,128,balanced,0.03822933385769526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,128,balanced,0.04071466624736786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,128,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,128,balanced,0.12628266215324402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,128,balanced,0.058245331048965454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,128,balanced,0.16722132762273154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,128,balanced,0.07286933561166127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,128,balanced,0.26874667406082153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,128,balanced,0.14476799964904785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,128,balanced,0.09340266386667888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,128,balanced,0.12807466586430868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,128,balanced,0.20057066281636557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,128,balanced,0.15963733196258545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,128,balanced,0.20597867170969644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,128,balanced,0.19156267245610556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,128,balanced,0.2921760082244873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,128,balanced,0.29127999146779376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,128,balanced,0.3715786536534627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,128,balanced,0.5394826730092367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,128,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,128,power_law_1.01,0.03283840119838714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,128,power_law_1.01,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,128,power_law_1.01,0.03412480056285858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,128,power_law_1.01,0.03489919900894165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,128,power_law_1.01,0.036396801471710205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,128,power_law_1.01,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,128,power_law_1.01,0.03583360016345978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,128,balanced,0.5534186760584513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,128,power_law_1.01,0.03676159977912903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,128,power_law_1.01,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,128,power_law_1.01,0.03768959939479828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,128,power_law_1.01,0.037254399061203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,128,power_law_1.01,0.038873600959777835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,128,power_law_1.01,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,128,power_law_1.01,0.04080640077590943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,128,power_law_1.01,0.041119998693466185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,128,power_law_1.01,0.041868799924850465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,128,power_law_1.01,0.043084800243377686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,128,power_law_1.01,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,128,power_law_1.01,0.04917759895324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,128,power_law_1.01,0.05161600112915039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,128,power_law_1.01,0.06172800064086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,128,power_law_1.01,0.06913920044898987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,128,power_law_1.01,0.09489279985427856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,128,power_law_1.01,0.11459840536117553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,128,power_law_1.01,0.14918400049209596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,128,power_law_1.01,0.18044159412384034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,128,power_law_1.01,0.24474239349365234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,128,power_law_1.01,0.36293759346008303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,128,power_law_1.01,0.696127986907959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,128,power_law_1.01,0.03411200046539307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,128,power_law_1.01,0.03159680068492889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,128,power_law_1.01,0.03292160034179688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,128,power_law_1.01,0.03431679904460907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,128,power_law_1.01,0.03389439880847931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,128,power_law_1.01,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,128,power_law_1.01,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,128,power_law_1.01,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,128,power_law_1.01,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,128,power_law_1.01,0.03699840009212494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,128,power_law_1.01,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,128,power_law_1.01,0.036236798763275145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,128,power_law_1.01,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,128,power_law_1.01,0.03904640078544617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,128,power_law_1.01,0.04170880019664765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,128,power_law_1.01,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,128,power_law_1.01,0.04252800047397613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,128,power_law_1.01,0.03162240087985992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,128,power_law_1.01,0.04568319916725159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,128,power_law_1.01,0.03163520097732544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,128,power_law_1.01,0.04775680005550385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,128,power_law_1.01,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,128,power_law_1.01,0.05235199928283692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,128,power_law_1.01,0.03410559892654419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,128,power_law_1.01,0.03367680013179779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,128,power_law_1.01,0.058310401439666745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,128,power_law_1.01,0.0350271999835968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,128,power_law_1.01,0.06933760046958923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,128,power_law_1.01,0.0356799989938736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,128,power_law_1.01,0.08120959997177124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,128,power_law_1.01,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,128,power_law_1.01,0.11978880167007447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,128,power_law_1.01,0.03596160113811493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,128,power_law_1.01,0.1488703966140747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,128,power_law_1.01,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,128,power_law_1.01,0.21412479877471924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,128,power_law_1.01,0.037049600481987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,128,power_law_1.01,0.2632895946502686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,128,power_law_1.01,0.03738879859447479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,128,power_law_1.01,0.3267263889312744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,128,power_law_1.01,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,128,power_law_1.01,0.5409791946411133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,128,power_law_1.01,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,128,power_law_1.01,0.01701759994029999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,128,power_law_1.01,0.040320000052452086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,128,power_law_1.01,1.1647680282592774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,128,power_law_1.01,0.018035200238227845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,128,power_law_1.01,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,128,power_law_1.01,0.018668800592422485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,128,power_law_1.01,0.04262399971485138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,128,power_law_1.01,0.02082560062408447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,128,power_law_1.01,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,128,power_law_1.01,0.03258239924907684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,128,power_law_1.01,0.0474368005990982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,128,power_law_1.01,0.02771199941635132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,128,power_law_1.01,0.05140479803085327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,128,power_law_1.01,0.02739199995994568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,128,power_law_1.01,0.057580798864364624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,128,power_law_1.01,0.02791680097579956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,128,power_law_1.01,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,128,power_law_1.01,0.026771199703216553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,128,power_law_1.01,0.07310079932212829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,128,power_law_1.01,0.027526399493217467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,128,power_law_1.01,0.0964352011680603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,128,power_law_1.01,0.028012800216674804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,128,power_law_1.01,0.02720640003681183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,128,power_law_1.01,0.11932799816131592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,128,power_law_1.01,0.027590399980545043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,128,power_law_1.01,0.15576959848403932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,128,power_law_1.01,0.027961599826812743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,128,power_law_1.01,0.19559040069580078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,128,power_law_1.01,0.030188798904418945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,128,power_law_1.01,0.2651456117630005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,128,power_law_1.2,0.03294079899787903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,128,power_law_1.01,0.031788799166679385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,128,power_law_1.2,0.03243519961833954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,128,power_law_1.01,0.3803839921951294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,128,power_law_1.01,0.031744000315666196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,128,power_law_1.2,0.03310079872608185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,128,power_law_1.01,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,128,power_law_1.01,0.7058815956115723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,128,power_law_1.2,0.03434880077838898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,128,power_law_1.01,0.03526400029659271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,128,power_law_1.2,0.03447679877281189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,128,power_law_1.01,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,128,power_law_1.2,0.03653759956359863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,128,power_law_1.01,0.040531200170516965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,128,power_law_1.2,0.03650560081005096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,128,power_law_1.01,0.04741120040416717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,128,power_law_1.2,0.03664000034332275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,128,power_law_1.01,0.054636800289154054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,128,power_law_1.2,0.03524479866027832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,128,power_law_1.01,0.06858879923820496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,128,power_law_1.2,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,128,power_law_1.01,0.08627840280532836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,128,power_law_1.2,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,128,power_law_1.01,0.11855360269546508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,128,power_law_1.2,0.03586559891700745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,128,power_law_1.2,0.03726719915866852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,128,power_law_1.01,0.1525823950767517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,128,power_law_1.2,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,128,power_law_1.01,0.1867136001586914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,128,power_law_1.2,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,128,power_law_1.01,0.2755840063095093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,128,power_law_1.2,0.04137600064277649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,128,power_law_1.2,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,128,power_law_1.01,0.5277120113372803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,128,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,128,power_law_1.2,0.04782719910144806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,128,power_law_1.2,0.054630398750305176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,128,power_law_1.2,0.05973119735717773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,128,power_law_1.2,0.07038080096244811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,128,power_law_1.2,0.08717439770698547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,128,power_law_1.2,0.11812479496002197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,128,power_law_1.2,0.14945919513702394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,128,power_law_1.2,0.19848959445953368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,128,power_law_1.2,0.26839680671691896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,128,power_law_1.2,0.3104127883911133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,128,power_law_1.2,0.5045567989349365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,128,power_law_1.2,0.9666815757751465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,128,power_law_1.2,0.03144319951534271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,128,power_law_1.2,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,128,power_law_1.2,0.031891199946403506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,128,power_law_1.2,0.034406399726867674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,128,power_law_1.2,0.0339711993932724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,128,power_law_1.2,0.03516800105571747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,128,power_law_1.2,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,128,power_law_1.2,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,128,power_law_1.2,0.03635199964046478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,128,power_law_1.2,0.036287999153137206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,128,power_law_1.2,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,128,power_law_1.2,0.037196800112724304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,128,power_law_1.2,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,128,power_law_1.2,0.03808000087738037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,128,power_law_1.2,0.0412992000579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,128,power_law_1.2,0.04208639860153198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,128,power_law_1.2,0.04281600117683411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,128,power_law_1.2,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,128,power_law_1.2,0.046316799521446225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,128,power_law_1.2,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,128,power_law_1.2,0.04992640018463135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,128,power_law_1.2,0.03272320032119751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,128,power_law_1.2,0.060288000106811526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,128,power_law_1.2,0.03348479866981506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,128,power_law_1.2,0.06447359919548035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,128,power_law_1.2,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,128,power_law_1.2,0.08196480274200439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,128,power_law_1.2,0.03516159951686859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,128,power_law_1.2,0.10008959770202637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,128,power_law_1.2,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,128,power_law_1.2,0.1457919955253601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,128,power_law_1.2,0.03603839874267578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,128,power_law_1.2,0.18519680500030516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,128,power_law_1.2,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,128,power_law_1.2,0.2589695930480957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,128,power_law_1.2,0.03696640133857727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,128,power_law_1.2,0.34206719398498536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,128,power_law_1.2,0.03735679984092712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,128,power_law_1.2,0.42046079635620115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,128,power_law_1.2,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,128,power_law_1.2,0.019180800020694732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,128,power_law_1.2,0.6559872150421142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,128,power_law_1.2,0.03834879994392395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,128,power_law_1.2,0.018016000092029572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,128,power_law_1.2,1.2382847785949707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,128,power_law_1.2,0.03738879859447479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,128,power_law_1.2,0.019167999923229217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,128,power_law_1.2,0.04182400107383728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,128,power_law_1.2,0.02067199945449829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,128,power_law_1.2,0.04126720130443573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,128,power_law_1.2,0.032179200649261476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,128,power_law_1.2,0.04254080057144165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,128,power_law_1.2,0.02752000093460083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,128,power_law_1.2,0.0460671991109848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,128,power_law_1.2,0.02728320062160492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,128,power_law_1.2,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,128,power_law_1.2,0.027910399436950683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,128,power_law_1.2,0.05612159967422485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,128,power_law_1.2,0.028038400411605834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,128,power_law_1.2,0.0633408010005951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,128,power_law_1.2,0.0274944007396698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,128,power_law_1.2,0.07624959945678711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,128,power_law_1.2,0.027923199534416198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,64,balanced,0.03203733265399933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,128,power_law_1.2,0.09076480269432068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,128,power_law_1.2,0.027564799785614012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,64,balanced,0.03322133421897888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,128,power_law_1.2,0.02810879945755005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,128,power_law_1.2,0.1204416036605835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,64,balanced,0.03319466610749563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,128,power_law_1.2,0.02839680016040802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,128,power_law_1.2,0.14709119796752929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,64,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,128,power_law_1.2,0.03018240034580231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,64,balanced,0.03646933287382126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,128,power_law_1.2,0.19466240406036378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,64,balanced,0.03839466720819473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,128,power_law_1.2,0.03086079955101013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,128,power_law_1.2,0.2605952024459839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,64,balanced,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,128,power_law_1.2,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,128,power_law_1.2,0.3294464111328125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,64,balanced,0.03713600089152654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,128,power_law_1.2,0.033990401029586795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,64,balanced,0.038805333276589714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,128,power_law_1.2,0.5136191844940186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,128,power_law_1.2,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,64,balanced,0.03942399968703588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,128,power_law_1.2,0.037254399061203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,128,power_law_1.2,0.9853568077087402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,64,balanced,0.03896533449490865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,128,power_law_1.2,0.041043201088905336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,64,balanced,0.03859733293453852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,128,power_law_1.2,0.047091200947761536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,64,balanced,0.0386613334218661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,128,power_law_1.2,0.058182400465011594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,64,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,128,power_law_1.2,0.07184640169143677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,64,balanced,0.04270400106906891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,128,power_law_1.2,0.09184640049934387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,64,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,128,power_law_1.2,0.12591999769210815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,64,balanced,0.04287466903527578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,128,power_law_1.2,0.16021759510040284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,64,balanced,0.042805333932240806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,128,power_law_1.2,0.19339519739151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,64,balanced,0.045040001471837364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,128,power_law_1.2,0.29173119068145753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,128,power_law_1.2,0.5520127773284912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,64,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,64,balanced,0.04695466657479604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,64,balanced,0.05012799799442291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,64,balanced,0.052255998055140175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,64,balanced,0.06150933106740316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,64,balanced,0.06659733255704244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,64,balanced,0.08548266688982646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,64,balanced,0.10146133104960124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,64,balanced,0.11782399813334148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,64,balanced,0.03206400076548258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,64,balanced,0.032986665765444435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,64,balanced,0.03297066688537598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,64,balanced,0.034202667574087776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,64,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,64,balanced,0.037061333656311035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,64,balanced,0.03679466744263967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,64,balanced,0.039034667114416756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,64,balanced,0.03827733298142751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,64,balanced,0.03867733230193456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,64,balanced,0.03862400104602178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,64,balanced,0.16686934232711792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,64,balanced,0.038378665844599404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,64,balanced,0.040709334115187325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,64,balanced,0.04045866678158442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,64,balanced,0.032085334261258446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,64,balanced,0.042853335539499916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,64,balanced,0.03435733417669932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,64,balanced,0.04443199932575226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,64,balanced,0.03451200077931086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,64,balanced,0.03621333340803782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,64,balanced,0.04470400015513102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,64,balanced,0.03646933287382126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,64,balanced,0.03851733356714249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,64,balanced,0.04621333380540212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,64,balanced,0.0391146664818128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,64,balanced,0.03847466657559077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,64,balanced,0.046522667010625206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,64,balanced,0.03879466652870178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,64,balanced,0.03849600007136663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,64,balanced,0.050741334756215416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,64,balanced,0.03865066667397817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,64,balanced,0.04035199930270513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,64,balanced,0.017690667261679966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,64,balanced,0.03849066545565923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,64,balanced,0.019930666933457058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,64,balanced,0.05269333223501841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,64,balanced,0.019989332805077236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,64,balanced,0.04035733391841253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,64,balanced,0.020202666521072388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,64,balanced,0.044026667873064675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,64,balanced,0.02179733415444692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,64,balanced,0.042463997999827065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,64,balanced,0.032655999064445496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,64,balanced,0.05909866591294607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,64,balanced,0.029450667401154835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,64,balanced,0.04429866870244344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,64,balanced,0.02881066749493281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,64,balanced,0.044682666659355164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,64,balanced,0.028229333460330963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,64,balanced,0.029504001140594482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,64,balanced,0.04811733464399973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,64,balanced,0.030266667405764263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,64,balanced,0.06162666777769724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,64,balanced,0.02985599885384242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,64,balanced,0.04866133133570353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,64,balanced,0.02978666623433431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,64,balanced,0.03018666555484136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,64,balanced,0.05151999990145365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,64,balanced,0.03092266619205475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,64,balanced,0.03202133377393087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,64,balanced,0.08018133540948232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,64,balanced,0.03200533241033554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,64,balanced,0.05665066838264465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,64,balanced,0.03446399917205175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,64,balanced,0.03687999894221624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,64,balanced,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,64,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,64,balanced,0.10422399640083313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,64,balanced,0.04469866553942362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,64,balanced,0.2918720046679179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,64,balanced,0.07429333527882893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,64,balanced,0.05065066615740458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,64,balanced,0.060959999759991966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,64,balanced,0.08506133159001668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,64,balanced,0.14470400412877402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,64,balanced,0.07890133559703827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,64,balanced,0.11199466387430827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,64,balanced,0.09813867012659709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,64,balanced,0.17617066701253256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,64,balanced,0.1368160049120585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,64,balanced,0.13237866759300232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,64,balanced,0.16826667388280234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,64,balanced,0.15863466262817383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,64,balanced,0.20901334285736084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,64,balanced,0.20179200172424316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,64,balanced,0.2258560061454773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,64,balanced,0.31089599927266437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,64,balanced,0.30744000275929767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,64,power_law_1.01,0.032595199346542356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,64,power_law_1.01,0.03164800107479095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,64,power_law_1.01,0.032627201080322264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,64,power_law_1.01,0.033888000249862674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,64,power_law_1.01,0.03457280099391937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,64,power_law_1.01,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,64,power_law_1.01,0.03635199964046478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,64,balanced,0.40805331865946454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,64,power_law_1.01,0.03546879887580871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,64,power_law_1.01,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,64,power_law_1.01,0.036627200245857236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,64,power_law_1.01,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,64,power_law_1.01,0.03759360015392303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,64,power_law_1.01,0.037452799081802365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,64,power_law_1.01,0.038361600041389464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,64,power_law_1.01,0.040608000755310056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,64,power_law_1.01,0.04089600145816803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,64,power_law_1.01,0.04154239892959595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,64,power_law_1.01,0.04317440092563629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,64,power_law_1.01,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,64,power_law_1.01,0.04631040096282959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,64,power_law_1.01,0.05262079834938049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,64,power_law_1.01,0.056883198022842404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,64,power_law_1.01,0.06535040140151978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,64,power_law_1.01,0.09344639778137206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,64,power_law_1.01,0.10468480587005616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,64,balanced,0.5861440102259318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,64,power_law_1.01,0.1475711941719055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,64,balanced,0.5734560092290243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,64,power_law_1.01,0.17964799404144288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,64,power_law_1.01,0.22366080284118653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,64,power_law_1.01,0.3311039924621582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,64,power_law_1.01,0.6527935981750488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,64,power_law_1.01,0.031692799925804135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,64,power_law_1.01,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,64,power_law_1.01,0.031462401151657104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,64,power_law_1.01,0.03296639919281006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,64,power_law_1.01,0.0342848002910614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,64,power_law_1.01,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,64,power_law_1.01,0.03529599905014038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,64,power_law_1.01,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,64,power_law_1.01,0.036287999153137206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,64,power_law_1.01,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,64,power_law_1.01,0.03703039884567261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,64,power_law_1.01,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,64,power_law_1.01,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,64,power_law_1.01,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,64,power_law_1.01,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,64,power_law_1.01,0.04239999949932098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,64,power_law_1.01,0.04251520037651062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,64,power_law_1.01,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,64,power_law_1.01,0.03282560110092163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,64,power_law_1.01,0.04612480103969574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,64,power_law_1.01,0.016864000260829924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,64,power_law_1.01,0.03252480030059814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,64,power_law_1.01,0.04836480021476745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,64,power_law_1.01,0.017664000391960144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,64,power_law_1.2,0.03296639919281006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,64,power_law_1.01,0.03543039858341217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,64,power_law_1.01,0.01759999990463257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,64,power_law_1.01,0.05320320129394531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,64,power_law_1.2,0.03251200020313263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,64,power_law_1.01,0.036134400963783266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,64,power_law_1.01,0.018291200697422027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,64,power_law_1.2,0.031839999556541446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,64,power_law_1.01,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,64,power_law_1.01,0.036985599994659425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,64,power_law_1.01,0.020870399475097657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,64,power_law_1.2,0.03416320085525513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,64,power_law_1.01,0.03723520040512085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,64,power_law_1.01,0.0709119975566864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,64,power_law_1.01,0.03154560029506683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,64,power_law_1.2,0.03438720107078552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,64,power_law_1.01,0.037273600697517395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,64,power_law_1.01,0.08448640108108521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,64,power_law_1.01,0.027647998929023743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,64,power_law_1.2,0.03498879969120026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,64,power_law_1.01,0.036934399604797365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,64,power_law_1.01,0.11842559576034546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,64,power_law_1.2,0.0359935998916626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,64,power_law_1.01,0.027507200837135315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,64,power_law_1.01,0.037248000502586365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,64,power_law_1.2,0.036908799409866334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,64,power_law_1.01,0.02736639976501465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,64,power_law_1.01,0.0378495991230011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,64,power_law_1.01,0.1525439977645874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,64,power_law_1.2,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,64,power_law_1.01,0.02714880108833313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,64,power_law_1.01,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,64,power_law_1.01,0.21031041145324708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,64,power_law_1.2,0.03660799860954285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,64,power_law_1.01,0.027315199375152588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,64,power_law_1.01,0.038796800374984744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,64,power_law_1.01,0.2705984115600586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,64,power_law_1.2,0.03668479919433594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,64,power_law_1.01,0.028121599555015565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,64,power_law_1.01,0.03937279880046844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,64,power_law_1.2,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,64,power_law_1.01,0.042368000745773314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,64,power_law_1.01,0.02748799920082092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,64,power_law_1.01,0.3197376012802124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,64,power_law_1.2,0.03861759901046753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,64,power_law_1.01,0.04336000084877014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,64,power_law_1.01,0.028288000822067262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,64,power_law_1.01,0.5131648063659668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,64,power_law_1.2,0.038464000821113585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,64,power_law_1.01,0.04374400079250336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,64,power_law_1.01,0.028960001468658448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,64,power_law_1.2,0.04092159867286682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,64,power_law_1.01,0.046540799736976626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,64,power_law_1.01,0.029849600791931153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,64,power_law_1.01,0.9985343933105468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,64,power_law_1.2,0.041657599806785586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,64,power_law_1.01,0.04965760111808777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,64,power_law_1.01,0.03059200048446655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,64,power_law_1.2,0.04209280014038086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,64,power_law_1.01,0.056985598802566526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,64,power_law_1.01,0.03290880024433136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,64,power_law_1.2,0.043750399351119997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,64,power_law_1.01,0.06238080263137817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,64,power_law_1.01,0.034771201014518735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,64,power_law_1.2,0.04472320079803467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,64,power_law_1.01,0.07123839855194092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,64,power_law_1.01,0.039628800749778745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,64,power_law_1.2,0.05162879824638367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,64,power_law_1.01,0.07874559760093688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,64,power_law_1.01,0.04164479970932007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,64,power_law_1.2,0.055327999591827395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,64,power_law_1.01,0.10030080080032348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,64,power_law_1.01,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,64,power_law_1.2,0.06900479793548583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,64,power_law_1.01,0.13982080221176146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,64,power_law_1.01,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,64,power_law_1.2,0.08179200291633607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,64,power_law_1.01,0.19511040449142455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,64,power_law_1.01,0.07379840016365051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,64,power_law_1.2,0.11196800470352172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,64,power_law_1.01,0.2178623914718628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,64,power_law_1.01,0.0902400016784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,64,power_law_1.2,0.15430400371551514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,64,power_law_1.01,0.25304958820343015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,64,power_law_1.01,0.12019840478897095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,64,power_law_1.2,0.19553920030593872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,64,power_law_1.01,0.4071487903594971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,64,power_law_1.01,0.15344640016555786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,64,power_law_1.2,0.23065600395202637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,64,power_law_1.01,0.18372479677200318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,64,power_law_1.01,0.9159487724304199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,64,power_law_1.2,0.3867199897766113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,64,power_law_1.01,0.28009600639343263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,64,power_law_1.2,0.6077760219573974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,64,power_law_1.01,0.5469376087188721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,64,power_law_1.2,1.190713596343994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,64,power_law_1.2,0.03160960078239441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,64,power_law_1.2,0.03020800054073334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,64,power_law_1.2,0.031488001346588135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,64,power_law_1.2,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,64,power_law_1.2,0.03445119857788086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,64,power_law_1.2,0.03495680093765259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,64,power_law_1.2,0.0350271999835968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,64,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,64,power_law_1.2,0.03583360016345978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,64,power_law_1.2,0.03633280098438263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,64,power_law_1.2,0.03798399865627289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,64,power_law_1.2,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,64,power_law_1.2,0.037215998768806456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,32,balanced,0.03190399954716364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,32,balanced,0.03242133309443792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,64,power_law_1.2,0.03857919871807099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,32,balanced,0.03266666581233343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,64,power_law_1.2,0.04182400107383728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,64,power_law_1.2,0.03256320059299469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,32,balanced,0.03482666611671448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,64,power_law_1.2,0.04338560104370117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,64,power_law_1.2,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,32,balanced,0.03730666637420654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,64,power_law_1.2,0.017280000448226928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,64,power_law_1.2,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,32,balanced,0.03678400069475174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,64,power_law_1.2,0.03344640135765076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,64,power_law_1.2,0.01849599927663803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,32,balanced,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,64,power_law_1.2,0.04655359983444214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,64,power_law_1.2,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,64,power_law_1.2,0.018105599284172057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,32,balanced,0.03859733293453852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,64,power_law_1.2,0.0487744003534317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,64,power_law_1.2,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,64,power_law_1.2,0.01905920058488846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,32,balanced,0.03719999889532725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,64,power_law_1.2,0.05882880091667175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,64,power_law_1.2,0.036748799681663516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,64,power_law_1.2,0.02120320051908493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,32,balanced,0.03913066784540812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,64,power_law_1.2,0.06537600159645081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,64,power_law_1.2,0.03694719970226288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,64,power_law_1.2,0.032416000962257385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,32,balanced,0.03894400099913279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,64,power_law_1.2,0.07982079982757569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,64,power_law_1.2,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,64,power_law_1.2,0.027046400308609008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,32,balanced,0.03879466652870178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,64,power_law_1.2,0.03754880130290985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,64,power_law_1.2,0.09960319995880126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,64,power_law_1.2,0.028966400027275085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,32,balanced,0.04043733328580856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,64,power_law_1.2,0.03790079951286316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,64,power_law_1.2,0.14027520418167114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,64,power_law_1.2,0.02802560031414032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,32,balanced,0.03884266565243403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,64,power_law_1.2,0.03800959885120392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,64,power_law_1.2,0.16967040300369263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,64,power_law_1.2,0.027430400252342224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,32,balanced,0.04104000081618627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,64,power_law_1.2,0.038150399923324585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,64,power_law_1.2,0.027596798539161683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,64,power_law_1.2,0.2974400043487549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,32,balanced,0.043247997760772705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,64,power_law_1.2,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,64,power_law_1.2,0.028435200452804565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,64,power_law_1.2,0.3771967887878418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,64,power_law_1.2,0.039878401160240176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,32,balanced,0.04251199960708618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,64,power_law_1.2,0.02739199995994568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,64,power_law_1.2,0.468723201751709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,64,power_law_1.2,0.04260480105876922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,64,power_law_1.2,0.02844800055027008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,32,balanced,0.04354133208592733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,64,power_law_1.2,0.8539327621459961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,64,power_law_1.2,0.042956799268722534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,64,power_law_1.2,0.02985599935054779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,64,power_law_1.2,0.04296320080757141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,64,power_law_1.2,0.02892799973487854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,32,balanced,0.04247466723124186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,64,power_law_1.2,1.7244991302490233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,64,power_law_1.2,0.04926080107688904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,64,power_law_1.2,0.030681601166725157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,64,power_law_1.2,0.05259519815444946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,64,power_law_1.2,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,32,balanced,0.04636266827583313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,64,power_law_1.2,0.05800960063934326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,64,power_law_1.2,0.03416320085525513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,64,power_law_1.2,0.062035202980041504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,64,power_law_1.2,0.03950720131397247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,32,balanced,0.04864533245563507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,64,power_law_1.2,0.08270080089569092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,64,power_law_1.2,0.04273279905319214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,64,power_law_1.2,0.09091839790344239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,64,power_law_1.2,0.04690560102462769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,64,power_law_1.2,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,64,power_law_1.2,0.13537919521331787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,32,balanced,0.05109866460164388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,64,power_law_1.2,0.07671040296554565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,64,power_law_1.2,0.18071680068969725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,64,power_law_1.2,0.09586560130119323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,64,power_law_1.2,0.2413952112197876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,64,power_law_1.2,0.13102079629898072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,64,power_law_1.2,0.3355007886886597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,64,power_law_1.2,0.1660416007041931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,64,power_law_1.2,0.41971840858459475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,32,balanced,0.05517866710821787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,64,power_law_1.2,0.20007679462432862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,64,power_law_1.2,0.7120192050933838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,64,power_law_1.2,0.3015360116958618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,64,power_law_1.2,1.3804224014282227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,64,power_law_1.2,0.5784192085266113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,32,balanced,0.06569066643714905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,32,balanced,0.07521600027879079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,32,balanced,0.09565866986910503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,32,balanced,0.11531733473141988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,32,balanced,0.1347040037314097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,32,balanced,0.029882666965325672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,32,balanced,0.0320266659061114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,32,balanced,0.031983998914559685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,32,balanced,0.034671999514102936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,32,balanced,0.03711466739575068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,32,balanced,0.03684266656637192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,32,balanced,0.03666666646798452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,32,balanced,0.039066667358080544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,32,balanced,0.03811733424663544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,32,balanced,0.03845866769552231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,32,balanced,0.038848000268141426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,32,balanced,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,32,balanced,0.04033066580692927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,32,balanced,0.04038399954636892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,32,balanced,0.04240000247955322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,32,balanced,0.04461333155632019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,32,balanced,0.044624000787734985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,32,balanced,0.04597333570321401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,32,power_law_1.01,0.0336896002292633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,32,balanced,0.04630400240421295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,32,power_law_1.01,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,32,power_law_1.01,0.03173120021820068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,32,balanced,0.05203199883302053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,32,power_law_1.01,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,32,power_law_1.01,0.03447679877281189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,32,power_law_1.01,0.036499199271202085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,32,balanced,0.05284800132115682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,32,power_law_1.01,0.03529599905014038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,32,power_law_1.01,0.033036801218986514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,32,power_law_1.01,0.03543039858341217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,32,power_law_1.01,0.03187200129032135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,32,power_law_1.01,0.0361407995223999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,32,balanced,0.06107733150323232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,32,power_law_1.01,0.032691198587417605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,32,power_law_1.01,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,32,power_law_1.01,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,32,power_law_1.01,0.03726719915866852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,32,power_law_1.01,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,32,power_law_1.01,0.03887999951839447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,32,power_law_1.01,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,32,power_law_1.01,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,32,balanced,0.06676800052324931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,32,power_law_1.01,0.03678719997406006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,32,power_law_1.01,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,32,power_law_1.01,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,32,power_law_1.01,0.040243199467658995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,32,power_law_1.01,0.03570559918880463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,32,power_law_1.01,0.04080640077590943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,32,power_law_1.01,0.037049600481987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,32,power_law_1.01,0.041791999340057374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,32,power_law_1.01,0.03754239976406097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,32,power_law_1.01,0.043750399351119997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,32,balanced,0.09294933080673218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,32,power_law_1.01,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,32,power_law_1.01,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,32,power_law_1.01,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,32,power_law_1.01,0.05081599950790405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,32,power_law_1.01,0.03865599930286408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,32,power_law_1.01,0.05249279737472534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,32,power_law_1.01,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,32,power_law_1.01,0.061868798732757566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,32,power_law_1.01,0.04336639940738678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,32,power_law_1.01,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,32,balanced,0.11559466520945232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,32,power_law_1.01,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,32,power_law_1.01,0.09433599710464477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,32,power_law_1.01,0.04700160026550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,32,power_law_1.01,0.10720000267028809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,32,power_law_1.01,0.04909439980983734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,32,power_law_1.01,0.13479039669036866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,32,power_law_1.01,0.0542464017868042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,32,power_law_1.01,0.1860159993171692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,32,power_law_1.01,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,32,power_law_1.01,0.2536128044128418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,32,power_law_1.01,0.07043200135231018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,32,power_law_1.01,0.3613375902175903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,32,power_law_1.01,0.08965759873390197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,32,balanced,0.15653866529464722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,32,power_law_1.01,0.6778560161590577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,32,power_law_1.01,0.13018879890441895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,32,power_law_1.01,0.1568063974380493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,32,power_law_1.01,0.23149440288543702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,32,power_law_1.01,0.26996479034423826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,32,power_law_1.01,0.36199679374694826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,32,power_law_1.01,0.5333183765411377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,32,power_law_1.01,1.0934656143188477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,32,balanced,0.19383466243743896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,32,balanced,0.23095466693242392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,32,balanced,0.34357865651448566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,32,balanced,0.19164800643920898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,32,balanced,0.6392746766408285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,32,power_law_1.2,0.03245440125465393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,32,power_law_1.2,0.03224320113658905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,32,power_law_1.2,0.03224320113658905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,32,power_law_1.2,0.033471998572349546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,32,power_law_1.2,0.03426559865474701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,32,power_law_1.2,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,32,power_law_1.2,0.03663359880447388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,32,power_law_1.2,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,32,power_law_1.2,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,32,power_law_1.2,0.036774399876594546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,32,power_law_1.2,0.031481599807739256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,32,power_law_1.2,0.036051198840141296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,32,power_law_1.2,0.036800000071525577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,32,power_law_1.2,0.03666560053825378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,32,power_law_1.2,0.03760640025138855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,32,power_law_1.2,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,32,power_law_1.2,0.041388800740242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,32,power_law_1.2,0.041382399201393125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,32,power_law_1.2,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,32,power_law_1.2,0.04362240135669708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,32,balanced,0.34147198994954425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,32,power_law_1.2,0.04785279929637909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,32,power_law_1.2,0.051820802688598636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,32,power_law_1.2,0.05553920269012451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,32,power_law_1.2,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,32,power_law_1.2,0.08860160112380981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,32,power_law_1.2,0.10705920457839965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,32,power_law_1.2,0.13559679985046386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,32,power_law_1.2,0.19579520225524902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,32,power_law_1.2,0.2477952003479004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,32,power_law_1.2,0.34155519008636476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,32,power_law_1.2,0.5612160205841065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,32,power_law_1.2,1.1902976036071777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,128,balanced,0.046821330984433494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,128,balanced,0.0489279975493749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,128,balanced,0.04822933177153269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,128,balanced,0.04845866560935974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,128,balanced,0.04937066634496053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,128,balanced,0.04924266537030538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,128,balanced,0.05065600077311198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,128,balanced,0.052970667680104576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,128,balanced,0.04951466619968414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,128,balanced,0.04996266464392344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,128,balanced,0.05116266508897146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,128,balanced,0.05081599950790405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,128,balanced,0.052517334620157875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,128,balanced,0.05118933320045471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,128,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,128,balanced,0.05505066613356272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,128,balanced,0.05266133447488149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,128,balanced,0.05343466500441233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,128,balanced,0.05219733218352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,128,balanced,0.05667200187842051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,128,balanced,0.05842133363087972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,128,balanced,0.06470400094985962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,128,balanced,0.07298133273919423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,128,balanced,0.03988266736268997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,128,balanced,0.08661866188049316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,128,balanced,0.04124800115823746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,128,balanced,0.042725334564844765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,128,balanced,0.043247997760772705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,128,balanced,0.043375998735427856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,128,balanced,0.045007998744646706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,128,balanced,0.042624001701672874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,128,balanced,0.045007998744646706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,128,balanced,0.04663999875386556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,128,balanced,0.0979253351688385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,128,balanced,0.044677332043647766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,128,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,128,balanced,0.04487466812133789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,128,balanced,0.04888000090916952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,128,balanced,0.046623999873797096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,128,balanced,0.0481279989083608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,128,balanced,0.050160000721613564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,128,balanced,0.05057600140571594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,32,power_law_1.2,0.03194240033626557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,128,balanced,0.1251146694024404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,32,power_law_1.2,0.033964800834655764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,128,balanced,0.052784000833829246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,32,power_law_1.2,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,128,balanced,0.05322133501370748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,32,power_law_1.2,0.03611519932746887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,32,power_law_1.2,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,128,balanced,0.04010133445262909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,32,power_law_1.2,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,128,balanced,0.06088533500830332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,128,balanced,0.04033066580692927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,128,balanced,0.040234667559464775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,32,power_law_1.2,0.03592320084571839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,128,balanced,0.041109333435694374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,32,power_law_1.2,0.03722879886627197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,128,balanced,0.06311999758084615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,128,balanced,0.042992000778516136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,32,power_law_1.2,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,128,balanced,0.0448586642742157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,32,power_law_1.2,0.03781760036945343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,128,balanced,0.044453332821528115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,128,balanced,0.15658666690190634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,128,balanced,0.07893333335717519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,128,balanced,0.045269335309664406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,32,power_law_1.2,0.03860479891300202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,128,balanced,0.04428799947102865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,32,power_law_1.2,0.039315199851989745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,128,balanced,0.045034666856129967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,32,power_law_1.2,0.04229759871959686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,128,balanced,0.04465599854787191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,32,power_law_1.2,0.04314239919185638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,128,balanced,0.09934399525324504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,128,balanced,0.04649066428343455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,32,power_law_1.2,0.044787201285362246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,128,balanced,0.04644800225893656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,32,power_law_1.2,0.048467200994491574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,128,balanced,0.046336000164349876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,32,power_law_1.2,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,128,balanced,0.04864533245563507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,32,power_law_1.2,0.06181120276451111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,128,balanced,0.1372106671333313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,32,power_law_1.2,0.06340479850769043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,128,balanced,0.04855466882387797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,32,power_law_1.2,0.08052480220794678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,128,balanced,0.05041066805521647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,32,power_law_1.2,0.09800320267677307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,128,balanced,0.051130667328834534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,128,balanced,0.1834719975789388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,32,power_law_1.2,0.1421183943748474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,128,balanced,0.1641759971777598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,32,power_law_1.2,0.19009920358657836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,128,balanced,0.0525439977645874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,32,power_law_1.2,0.2921087980270386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,128,balanced,0.05668266614278158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,32,power_law_1.2,0.43695359230041503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,32,power_law_1.2,0.4254144191741943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,128,balanced,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,32,power_law_1.2,0.8103679656982422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,32,power_law_1.2,1.4959872245788575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,128,balanced,0.22099733352661133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,128,balanced,0.07153599957625072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,128,balanced,0.07958399752775829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,128,balanced,0.2595733404159546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,128,balanced,0.282586673895518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,128,balanced,0.1039680043856303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,128,balanced,0.12133333086967468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,128,balanced,0.3381066719690959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,128,balanced,0.15685866276423135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,128,balanced,0.1937546730041504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,128,balanced,0.5093813339869181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,128,balanced,0.47383999824523926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,128,balanced,0.22957332928975424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,128,balanced,0.9645013014475504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,128,balanced,0.3405439853668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,128,balanced,0.020453333854675293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,128,balanced,0.02086399992307027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,128,balanced,0.023183998962243397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,128,balanced,0.026309333741664886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,128,balanced,0.04267199834187826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,128,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,128,balanced,0.03484266748030981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,128,balanced,0.034661332766215004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,128,balanced,0.03659733384847641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,128,balanced,0.036848001182079315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,128,balanced,0.037045332292715706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,128,balanced,0.03691199918588003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,128,balanced,0.035930665830771126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,128,balanced,0.03701333453257879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,128,balanced,0.040709334115187325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,128,balanced,0.04051200052102407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,128,balanced,0.038704000413417816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,128,balanced,0.041722665230433144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,128,balanced,0.04413333535194397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,128,balanced,0.049738665421803795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,128,balanced,0.06098666787147522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,128,balanced,0.07384000221888225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,128,balanced,0.09128533800443013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,128,balanced,0.12474133570988973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,128,balanced,0.6265920003255209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,128,balanced,0.15681599577267966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,128,power_law_1.01,0.045721599459648134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,128,power_law_1.01,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,128,power_law_1.01,0.048256000876426695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,128,balanced,0.2209119995435079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,128,power_law_1.01,0.0494592010974884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,128,power_law_1.01,0.04791040122509003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,128,power_law_1.01,0.04867199957370758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,128,power_law_1.01,0.048198398947715757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,128,power_law_1.01,0.0483711987733841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,128,power_law_1.01,0.046463999152183535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,128,power_law_1.01,0.046393600106239316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,128,power_law_1.01,0.047167998552322385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,128,power_law_1.01,0.04609920084476471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,128,power_law_1.01,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,128,power_law_1.01,0.045587199926376346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,128,power_law_1.01,0.04981119930744171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,128,balanced,0.28428266445795697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,128,power_law_1.01,0.05193600058555603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,128,power_law_1.01,0.05103359818458557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,128,power_law_1.01,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,128,power_law_1.01,0.0673088014125824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,128,power_law_1.01,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,128,power_law_1.01,0.08477439880371093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,128,power_law_1.01,0.11341439485549927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,128,power_law_1.01,0.13319040536880494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,128,power_law_1.01,0.1707136034965515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,128,power_law_1.01,0.21571199893951415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,128,power_law_1.01,0.3432640075683594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,128,power_law_1.01,0.4048128128051758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,128,power_law_1.01,0.5306496143341064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,128,balanced,0.34785600503285724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,128,power_law_1.01,0.8734463691711426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,128,power_law_1.01,0.03782399892807007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,128,power_law_1.01,1.7609535217285157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,128,power_law_1.01,0.039052799344062805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,128,power_law_1.01,0.038150399923324585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,128,power_law_1.01,0.040703999996185306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,128,power_law_1.01,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,128,power_law_1.01,0.04252800047397613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,128,power_law_1.01,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,128,power_law_1.01,0.042259201407432556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,128,power_law_1.01,0.04238080084323883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,128,power_law_1.01,0.04259839951992035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,128,power_law_1.01,0.04184960126876831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,128,power_law_1.01,0.04222719967365265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,128,power_law_1.01,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,128,power_law_1.01,0.04299519956111908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,128,power_law_1.01,0.04780159890651703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,128,power_law_1.01,0.04909439980983734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,128,power_law_1.01,0.048742398619651794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,128,balanced,0.5388053258260092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,128,power_law_1.01,0.05472000241279602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,128,power_law_1.01,0.05909759998321533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,128,power_law_1.01,0.07087360024452209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,128,power_law_1.01,0.08316159844398499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,128,power_law_1.01,0.11498240232467652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,128,power_law_1.01,0.1435263991355896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,128,power_law_1.01,0.19178240299224852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,128,power_law_1.01,0.23562240600585938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,128,power_law_1.01,0.3549247980117798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,128,power_law_1.01,0.4719808101654053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,128,power_law_1.01,0.5837823867797851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,128,power_law_1.01,0.9327808380126953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,128,power_law_1.01,1.8715839385986328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,128,power_law_1.01,0.03790720105171204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,128,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,128,power_law_1.01,0.0377344012260437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,128,power_law_1.01,0.04051199853420258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,128,power_law_1.01,0.040889599919319154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,128,power_law_1.01,0.042387199401855466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,128,balanced,1.0452533562978108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,128,power_law_1.01,0.04035199880599975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,128,power_law_1.01,0.04217599928379059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,128,power_law_1.01,0.041407999396324155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,128,power_law_1.01,0.04118399918079376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,128,power_law_1.01,0.042444801330566405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,128,power_law_1.01,0.0427839994430542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,128,power_law_1.01,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,128,power_law_1.01,0.0426367998123169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,128,power_law_1.01,0.04591360092163086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,128,power_law_1.01,0.04636160135269165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,128,power_law_1.01,0.051635199785232545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,128,power_law_1.01,0.05369600057601929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.01,0.059628802537918094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.01,0.0656063973903656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.01,0.07617920041084289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.01,0.0972544014453888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.01,0.11755520105361938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.01,0.15548800230026244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.01,0.19339519739151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.01,0.2723072052001953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.01,0.3461888074874878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,128,power_law_1.01,0.43043198585510256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,128,power_law_1.01,0.7104959964752198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,128,power_law_1.01,1.4013440132141113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,128,power_law_1.01,0.018624000251293182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,128,power_law_1.01,0.01910399943590164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,128,power_law_1.01,0.020134399831295013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,128,power_law_1.01,0.02393600046634674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,128,power_law_1.01,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,128,power_law_1.01,0.033580800890922545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,128,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,128,power_law_1.01,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,128,power_law_1.01,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,128,power_law_1.01,0.03346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,128,power_law_1.01,0.034745600819587705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,128,power_law_1.01,0.03548159897327423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,128,power_law_1.01,0.034220799803733826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,128,power_law_1.01,0.034355199337005614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,128,power_law_1.01,0.035897600650787356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,128,power_law_1.01,0.03678719997406006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,128,power_law_1.01,0.03731839954853058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,128,power_law_1.01,0.03845120072364807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.01,0.04264959990978241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.01,0.04901120066642761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.01,0.056883198022842404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.01,0.06999040246009827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.01,0.08428159952163697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.01,0.11784319877624512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.01,0.14790400266647338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.01,0.2075968027114868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,128,power_law_1.2,0.048492801189422605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,128,power_law_1.2,0.048102399706840514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.01,0.262175989151001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,128,power_law_1.2,0.04791040122509003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,128,power_law_1.01,0.32851200103759765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,128,power_law_1.2,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,128,power_law_1.01,0.49496960639953613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,128,power_law_1.2,0.04837760031223297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,128,power_law_1.2,0.049292799830436704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,128,power_law_1.01,0.9886143684387207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,128,power_law_1.2,0.04989440143108368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,128,power_law_1.2,0.04928640127182007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,128,power_law_1.2,0.048179200291633605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,128,power_law_1.2,0.04695680141448975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,128,power_law_1.2,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,128,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,128,power_law_1.2,0.04588800072669983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,128,power_law_1.2,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,128,power_law_1.2,0.052691197395324706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,128,power_law_1.2,0.05013120174407959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,128,power_law_1.2,0.056390398740768434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,128,power_law_1.2,0.06021760106086731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,128,power_law_1.2,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,128,power_law_1.2,0.08765439987182617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,128,power_law_1.2,0.09952639937400817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,128,power_law_1.2,0.14124799966812135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,128,power_law_1.2,0.1775488018989563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,128,power_law_1.2,0.2561791896820068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,128,power_law_1.2,0.3231359958648682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,128,power_law_1.2,0.48521599769592283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,128,power_law_1.2,0.6269631862640381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,128,power_law_1.2,0.7775360107421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,128,power_law_1.2,1.2468992233276368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,128,power_law_1.2,2.460422325134277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,128,power_law_1.2,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,128,power_law_1.2,0.038899201154708865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,128,power_law_1.2,0.04074240028858185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,128,power_law_1.2,0.04136320054531097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,128,power_law_1.2,0.041894400119781496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,128,power_law_1.2,0.04291839897632599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,128,power_law_1.2,0.042694398760795595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,128,power_law_1.2,0.04220159947872162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,128,power_law_1.2,0.042208001017570496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,128,power_law_1.2,0.04268800020217896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,128,power_law_1.2,0.042559999227523806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,128,power_law_1.2,0.042054399847984314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,128,power_law_1.2,0.043532800674438474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,128,power_law_1.2,0.04529280066490173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,128,power_law_1.2,0.04898560047149658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,128,power_law_1.2,0.04870400130748749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,128,power_law_1.2,0.05296639800071716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,128,power_law_1.2,0.0564736008644104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,128,power_law_1.2,0.06792320013046264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,128,power_law_1.2,0.07834879755973816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,128,power_law_1.2,0.09772800207138062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,128,power_law_1.2,0.1444607973098755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,128,power_law_1.2,0.1665343999862671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,128,power_law_1.2,0.25027201175689695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,128,power_law_1.2,0.30101120471954346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,128,power_law_1.2,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,128,power_law_1.2,0.4468095779418945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,128,power_law_1.2,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,128,power_law_1.2,0.6101312160491943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,128,power_law_1.2,0.03893760144710541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,128,power_law_1.2,0.7410111904144288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,128,power_law_1.2,0.04056319892406464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,128,power_law_1.2,0.04099839925765991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,128,power_law_1.2,1.1555392265319824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,128,power_law_1.2,0.04224640130996704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,128,power_law_1.2,2.3353471755981445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,128,power_law_1.2,0.041791999340057374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,128,power_law_1.2,0.04230400025844574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,128,power_law_1.2,0.04181120097637177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,128,power_law_1.2,0.04323840141296387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,128,power_law_1.2,0.04154880046844482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,128,power_law_1.2,0.041407999396324155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,128,power_law_1.2,0.04177919924259186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,128,power_law_1.2,0.042822399735450746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,128,power_law_1.2,0.049209600687026976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,128,power_law_1.2,0.04826239943504333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,128,power_law_1.2,0.05121920108795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,128,power_law_1.2,0.06025599837303162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.2,0.06383360028266907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.2,0.07434880137443542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.2,0.08343679904937744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.2,0.1206976056098938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.2,0.1338047981262207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.2,0.2044287919998169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.2,0.2555840015411377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.2,0.3797440052032471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.2,0.4855807781219482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,128,power_law_1.2,0.6031487941741943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,128,power_law_1.2,0.9394495964050293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,128,power_law_1.2,1.8324544906616211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,128,power_law_1.2,0.018969599902629853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,128,power_law_1.2,0.019046400487422944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,128,power_law_1.2,0.020006400346755982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,128,power_law_1.2,0.02366719990968704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,128,power_law_1.2,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,128,power_law_1.2,0.032671999931335446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,128,power_law_1.2,0.033471998572349546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,128,power_law_1.2,0.03315199911594391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,128,power_law_1.2,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,128,power_law_1.2,0.03335680067539215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,128,power_law_1.2,0.03447679877281189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,128,power_law_1.2,0.03485440015792847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,128,power_law_1.2,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,128,power_law_1.2,0.032716798782348636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,128,power_law_1.2,0.036447998881340024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,128,power_law_1.2,0.03783040046691895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,128,power_law_1.2,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,128,power_law_1.2,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.2,0.04170239865779877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.2,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.2,0.057529598474502563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.2,0.07283200025558471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.2,0.0892416000366211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.2,0.12323839664459228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.2,0.1551424026489258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.2,0.21877760887145997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.2,0.28250880241394044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,128,power_law_1.2,0.3457024097442627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,128,power_law_1.2,0.538150405883789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,128,power_law_1.2,1.0441920280456543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,64,balanced,0.03993066648642222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,64,balanced,0.040976000328858696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,64,balanced,0.04118400067090988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,64,balanced,0.04257600009441376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,64,balanced,0.04475200176239014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,64,balanced,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,64,balanced,0.047210668524106346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,64,balanced,0.04656533400217692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,64,balanced,0.046336000164349876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,64,balanced,0.05097599824269613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,64,balanced,0.046538665890693665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,64,balanced,0.04667200148105621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,64,balanced,0.05494933327039083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,64,balanced,0.04625066618124644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,64,balanced,0.05315199991067251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,64,balanced,0.05095999936262766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,64,balanced,0.052576000491778054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,64,balanced,0.05197866757710775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,64,balanced,0.052602668603261314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,64,balanced,0.05477866530418396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,64,balanced,0.058634668588638306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,64,balanced,0.06778133412202199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,64,balanced,0.03631466627120972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,64,balanced,0.07344533503055573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,64,balanced,0.03811733424663544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,64,balanced,0.03928533444801966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,64,balanced,0.04049599915742874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,64,balanced,0.04244266450405121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,64,balanced,0.043338666359583534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,64,balanced,0.09099733829498291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,64,balanced,0.04443199932575226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,64,balanced,0.04251733422279358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,64,balanced,0.04284266630808512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,64,balanced,0.04304533203442892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,64,balanced,0.04491200049718221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,64,balanced,0.0462719996770223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,64,balanced,0.10469333330790202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,64,balanced,0.044922664761543274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,64,balanced,0.044549331068992615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,64,balanced,0.048991998036702476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,64,balanced,0.048538664976755776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,64,balanced,0.05061866839726766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,64,balanced,0.05288533369700114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,64,balanced,0.05682133138179779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,64,balanced,0.1301866670449575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,64,balanced,0.036176001032193504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,64,balanced,0.03681600093841553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,64,balanced,0.058778668443361916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,64,balanced,0.03738666574160258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,64,balanced,0.038975998759269714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,64,balanced,0.04306666553020477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,64,balanced,0.0665226678053538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,64,balanced,0.04247466723124186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,64,balanced,0.042453333735466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,64,balanced,0.043920000394185386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,64,balanced,0.04600533346335093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,64,balanced,0.08109333117802937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,64,balanced,0.042549331982930504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,64,balanced,0.044437333941459656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,64,balanced,0.0431573341290156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,64,balanced,0.16426133116086325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,64,balanced,0.046384001771608986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,64,balanced,0.046575998266537987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,64,balanced,0.10670933127403259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,64,balanced,0.04900800188382467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,64,balanced,0.048245335618654885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,64,balanced,0.050613333781560264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,64,balanced,0.051407997806866966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,64,balanced,0.1413386662801107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,64,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,64,balanced,0.059402664502461754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,64,balanced,0.06062399844328562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,64,balanced,0.18913066387176514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,64,balanced,0.16967467466990152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,64,balanced,0.07302933434645335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,64,balanced,0.08165866633256276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,64,balanced,0.10709866881370544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,64,balanced,0.22930665810902914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,64,balanced,0.020026666422684986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,64,balanced,0.020346666375796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,64,balanced,0.022543999056021374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,64,balanced,0.021914665897687275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,64,balanced,0.12619733810424805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,64,balanced,0.27220799525578815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,64,balanced,0.024282666544119518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,64,balanced,0.04119466741879781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,64,balanced,0.03489600121974945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,64,balanced,0.03412266572316488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,64,balanced,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,64,balanced,0.034661332766215004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,64,balanced,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,64,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,64,balanced,0.0342399999499321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,64,balanced,0.29203200340270996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,64,balanced,0.16484799981117249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,64,balanced,0.03454400102297465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,64,balanced,0.03915199885765711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,64,balanced,0.03863999992609024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,64,balanced,0.03847466657559077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,64,balanced,0.042821332812309265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,64,balanced,0.046154667933781944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,64,balanced,0.0491839994986852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,64,balanced,0.20363734165827432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,64,balanced,0.0628053347269694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,64,balanced,0.07678399980068207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,64,balanced,0.35326401392618817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,64,balanced,0.0953546663125356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,64,balanced,0.12576533357302347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,64,balanced,0.2407360076904297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,64,balanced,0.5003626743952433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,64,balanced,0.16127999623616537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,64,balanced,0.22409600019454956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,64,balanced,0.5294719934463501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,64,balanced,0.3574399948120117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,64,balanced,0.2892586588859558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,64,balanced,0.354751984278361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,64,balanced,0.5555146535237631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,64,balanced,0.6625866492589315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,64,balanced,1.0060053666432698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,64,power_law_1.01,0.04480000138282776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,64,power_law_1.01,0.040992000699043275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,64,power_law_1.01,0.0411327987909317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,64,balanced,1.078378677368164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,64,power_law_1.01,0.0441536009311676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,64,power_law_1.01,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,64,power_law_1.01,0.04477440118789673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,64,power_law_1.01,0.047014400362968445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,64,power_law_1.01,0.04613119959831238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,64,power_law_1.01,0.04607360064983368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,64,power_law_1.01,0.04611839950084686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,64,power_law_1.01,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,64,power_law_1.01,0.046463999152183535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,64,power_law_1.01,0.046342399716377256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,64,power_law_1.01,0.04562560021877289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,64,power_law_1.01,0.04816640019416809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,64,power_law_1.01,0.04984320104122162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,64,power_law_1.01,0.051481598615646364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,64,power_law_1.01,0.05434240102767944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,64,power_law_1.01,0.060659199953079224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,64,power_law_1.01,0.07018240094184876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,64,power_law_1.01,0.07489920258522034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,64,power_law_1.01,0.10142079591751099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,64,power_law_1.01,0.12461440563201905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,64,power_law_1.01,0.14924800395965576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,64,power_law_1.01,0.17895040512084961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,64,power_law_1.01,0.26739840507507323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,64,power_law_1.01,0.35825281143188475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,64,power_law_1.01,0.4550528049468994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,64,power_law_1.01,0.7107264041900635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,64,power_law_1.01,1.2302783966064452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,64,power_law_1.01,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,64,power_law_1.01,0.03781760036945343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,64,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,64,power_law_1.01,0.0400191992521286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,64,power_law_1.01,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,64,power_law_1.01,0.04017919898033142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,64,power_law_1.01,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,64,power_law_1.01,0.0406143993139267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,64,power_law_1.01,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,64,power_law_1.01,0.04056319892406464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,64,power_law_1.01,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,64,power_law_1.01,0.04116480052471161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,64,power_law_1.01,0.03978880047798157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,64,power_law_1.01,0.04088320136070252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,64,power_law_1.01,0.0411327987909317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,64,power_law_1.01,0.04216960072517395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,64,power_law_1.01,0.04225279986858368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,64,power_law_1.01,0.04106239974498749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,64,power_law_1.01,0.041075199842453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,64,power_law_1.01,0.04296320080757141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,64,power_law_1.01,0.04380159974098206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,64,power_law_1.01,0.04245760142803192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,64,power_law_1.01,0.04348799884319306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,64,power_law_1.01,0.043584001064300534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,64,power_law_1.01,0.04702720046043396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,64,power_law_1.01,0.047788798809051514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,64,power_law_1.01,0.04968959987163544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,64,power_law_1.01,0.055852800607681274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.01,0.05862399935722351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.01,0.06592640280723572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.01,0.07697280049324036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.01,0.0944320023059845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.01,0.11815680265426635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.01,0.14924800395965576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.01,0.18854399919509887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,64,power_law_1.01,0.01817599982023239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.01,0.24771199226379395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,64,power_law_1.01,0.018624000251293182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,64,power_law_1.01,0.018956799805164338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.01,0.3381311893463135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,64,power_law_1.01,0.020134399831295013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,64,power_law_1.01,0.39978880882263185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,64,power_law_1.01,0.02234880030155182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,64,power_law_1.01,0.6692800045013427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,64,power_law_1.01,0.039705601334571836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,64,power_law_1.01,0.03289600014686585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,64,power_law_1.01,1.3246399879455566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,64,power_law_1.01,0.03351680040359497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,64,power_law_1.01,0.033792001008987424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,64,power_law_1.01,0.03353599905967712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,64,power_law_1.01,0.033452799916267394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,64,power_law_1.01,0.03351680040359497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,64,power_law_1.01,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,64,power_law_1.01,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,64,power_law_1.01,0.03363839983940124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,64,power_law_1.01,0.037363201379776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,64,power_law_1.01,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,64,power_law_1.01,0.03894400000572205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.01,0.0436928004026413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.01,0.04803839921951294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.01,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.01,0.07189120054244995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.01,0.08520320057868958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.01,0.11089279651641845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.01,0.13896960020065308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.01,0.19906560182571412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.01,0.2637631893157959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,64,power_law_1.01,0.32366719245910647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,64,power_law_1.01,0.481990385055542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,64,power_law_1.01,0.939891242980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,64,power_law_1.2,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,64,power_law_1.2,0.040403199195861814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,64,power_law_1.2,0.04264959990978241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,64,power_law_1.2,0.04387840032577515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,64,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,64,power_law_1.2,0.046393600106239316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,64,power_law_1.2,0.04616959989070892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,64,power_law_1.2,0.04750080108642578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,64,power_law_1.2,0.04425599873065948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,64,power_law_1.2,0.04657920002937317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,64,power_law_1.2,0.04645760059356689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,64,power_law_1.2,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,64,power_law_1.2,0.044633600115776065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,64,power_law_1.2,0.04652160108089447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,64,power_law_1.2,0.051744002103805545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,64,power_law_1.2,0.052339202165603636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,64,power_law_1.2,0.05551999807357788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,64,power_law_1.2,0.05883520245552063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,64,power_law_1.2,0.06575359702110291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,64,power_law_1.2,0.08038399815559387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,64,power_law_1.2,0.09225599765777588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,64,power_law_1.2,0.12377599477767945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,64,power_law_1.2,0.13953280448913574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,64,power_law_1.2,0.19298559427261353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,64,power_law_1.2,0.23319680690765382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,64,power_law_1.2,0.41753602027893066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,64,power_law_1.2,0.6207104206085206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,64,power_law_1.2,0.6417600154876709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,64,power_law_1.2,1.2444928169250489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,64,power_law_1.2,2.7229888916015623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,64,power_law_1.01,0.04256640076637268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,64,power_law_1.01,0.042719998955726625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,64,power_law_1.01,0.04323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,64,power_law_1.01,0.04359680116176605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,64,power_law_1.01,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,64,power_law_1.01,0.04707199931144714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,64,power_law_1.01,0.05023360252380371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,64,power_law_1.01,0.05451520085334778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,64,power_law_1.01,0.060038399696350095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,64,power_law_1.2,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,64,power_law_1.01,0.07139840126037597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,64,power_law_1.2,0.0361407995223999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,64,power_law_1.01,0.08843520283699036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,64,power_law_1.2,0.038924801349639895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,64,power_law_1.01,0.11224960088729859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,64,power_law_1.2,0.039084801077842714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,64,power_law_1.01,0.1344063997268677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,64,power_law_1.2,0.039552000164985654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,64,power_law_1.01,0.1864192008972168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,64,power_law_1.2,0.04179840087890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,64,power_law_1.01,0.2340991973876953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,64,power_law_1.2,0.04110719859600067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,64,power_law_1.01,0.3351167917251587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,64,power_law_1.2,0.04158720076084137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,64,power_law_1.01,0.4944255828857422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,64,power_law_1.2,0.04169600009918213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,64,power_law_1.01,0.5872128009796143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,64,power_law_1.2,0.042694398760795595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,64,power_law_1.01,0.8648192405700683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,64,power_law_1.2,0.04259200096130371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,64,power_law_1.2,0.04567680060863495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,64,power_law_1.01,1.6975679397583008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,64,power_law_1.2,0.044095999002456664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,64,power_law_1.2,0.044300800561904906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,64,power_law_1.2,0.04764800071716309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,64,power_law_1.2,0.05004159808158874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,64,power_law_1.2,0.05278080105781555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,64,power_law_1.2,0.05873280167579651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,64,power_law_1.2,0.0629696011543274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,64,power_law_1.2,0.08042240142822266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,64,power_law_1.2,0.09595519900321961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,64,power_law_1.2,0.03857280015945434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,64,power_law_1.2,0.1372032046318054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,64,power_law_1.2,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,64,power_law_1.2,0.15826560258865358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,64,power_law_1.2,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,64,power_law_1.2,0.22716801166534423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,64,power_law_1.2,0.0395904004573822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,64,power_law_1.2,0.3003135919570923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,64,power_law_1.2,0.04160000085830688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,64,power_law_1.2,0.4987520217895508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,64,power_law_1.2,0.04206719994544983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,64,power_law_1.2,0.6037248134613037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,64,power_law_1.2,0.04245760142803192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,64,power_law_1.2,0.04073599874973297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,64,power_law_1.2,0.749132776260376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,64,power_law_1.2,0.04131200015544891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,64,power_law_1.2,1.3937408447265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,64,power_law_1.2,0.04128639996051788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,64,power_law_1.2,2.713190460205078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,64,power_law_1.2,0.043647998571395875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,64,power_law_1.2,0.043942400813102724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,64,power_law_1.2,0.043424001336097716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,64,power_law_1.2,0.045158401131629944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,64,power_law_1.2,0.0493120014667511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,64,power_law_1.2,0.04962559938430786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,64,power_law_1.2,0.05303040146827698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,64,power_law_1.2,0.05737599730491638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.2,0.06305919885635376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.2,0.0744704008102417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.2,0.08458880186080933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.2,0.11630719900131226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.2,0.1361791968345642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.2,0.17399040460586548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.2,0.24990720748901368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.2,0.3724800109863281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.2,0.5212160110473633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,64,power_law_1.2,0.530790376663208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,64,power_law_1.2,1.0526528358459473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,64,power_law_1.2,2.3197439193725584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,64,power_law_1.2,0.019276799261569978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,64,power_law_1.2,0.01940480023622513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,64,power_law_1.2,0.01974399983882904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,64,power_law_1.2,0.020006400346755982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,64,power_law_1.2,0.02396800071001053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,64,power_law_1.2,0.03909119963645935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,64,power_law_1.2,0.033555200695991515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,64,power_law_1.2,0.033632001280784606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,64,power_law_1.2,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,64,power_law_1.2,0.032364800572395325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,64,power_law_1.2,0.03276160061359405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,64,power_law_1.2,0.03335680067539215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,64,power_law_1.2,0.033632001280784606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,64,power_law_1.2,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,64,power_law_1.2,0.03375999927520752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,64,power_law_1.2,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,64,power_law_1.2,0.03759360015392303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,64,power_law_1.2,0.04028800129890442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.2,0.043782401084899905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.2,0.04839679896831513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.2,0.061612802743911746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.2,0.07493119835853576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.2,0.09184640049934387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.2,0.11841280460357666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.2,0.14965759515762328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.2,0.21607038974761963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.2,0.2770495891571045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,64,power_law_1.2,0.3492095947265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,64,power_law_1.2,0.5362304210662842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,64,power_law_1.2,1.0689791679382323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,32,balanced,0.03823466598987579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,32,balanced,0.038933334251244865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,32,balanced,0.03905066599448522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,32,balanced,0.04292800029118856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,32,balanced,0.044581333796183266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,32,balanced,0.048122664292653404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,32,balanced,0.047184000412623085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,32,balanced,0.04634666442871094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,32,balanced,0.05096533397833506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,32,balanced,0.04811733464399973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,32,balanced,0.04699199895064036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,32,balanced,0.04851733148097992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,32,balanced,0.04693866769472758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,32,balanced,0.04607999821503957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,32,balanced,0.05091733237107595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,32,balanced,0.048725331823031105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,32,balanced,0.05080533524354299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,32,balanced,0.051589335004488625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,32,balanced,0.05500266452630361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,32,balanced,0.056176001826922096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,32,balanced,0.05880000193913778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,32,balanced,0.03559466699759165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,32,balanced,0.036533333361148834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,32,balanced,0.03637866675853729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,32,balanced,0.03896533449490865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,32,balanced,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,32,balanced,0.07155733307202657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,32,balanced,0.0432640016078949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,32,balanced,0.04234666625658671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,32,balanced,0.044906665881474815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,32,balanced,0.04455466568470001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,32,balanced,0.08027199904123943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,32,balanced,0.042805333932240806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,32,balanced,0.04483200112978617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,32,balanced,0.04465066889921824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,32,balanced,0.04660266637802124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,32,balanced,0.046207999189694725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,32,balanced,0.048154667019844055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,32,balanced,0.09833066662152608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,32,balanced,0.050527999798456825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,32,balanced,0.050383999943733215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,32,balanced,0.05468266705671946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,32,balanced,0.03594133257865906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,32,balanced,0.03702399879693985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,32,balanced,0.05522666871547699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,32,balanced,0.11256532867749532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,32,balanced,0.03642666588226954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,32,balanced,0.04043733328580856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,32,balanced,0.04506133496761322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,32,balanced,0.06287999947865804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,32,balanced,0.04484800000985464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,32,balanced,0.04448533554871877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,32,balanced,0.06866666674613953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,32,balanced,0.046154667933781944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,32,balanced,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,32,balanced,0.04500266909599304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,32,balanced,0.04675200084845225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,32,balanced,0.08915733297665913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,32,balanced,0.04664533336957296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,32,balanced,0.14269866545995077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,32,balanced,0.0487360010544459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,32,balanced,0.04855466882387797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,32,balanced,0.05027199784914652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,32,balanced,0.11198400457700093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,32,balanced,0.05128000179926554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,32,balanced,0.052709331115086876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,32,balanced,0.054325332244237266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,32,balanced,0.05477866530418396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,32,balanced,0.14827733238538107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,32,balanced,0.06252799928188324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,32,balanced,0.17909866571426392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,32,balanced,0.06517866750558217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,32,balanced,0.17975467443466187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,32,balanced,0.08038400113582611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,32,balanced,0.09105599919954936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,32,balanced,0.24596265951792398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,32,balanced,0.11692800124486287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,32,balanced,0.21192532777786255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,32,balanced,0.1381119986375173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,32,balanced,0.01937599976857503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,32,balanced,0.022064000368118286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,32,balanced,0.020128000527620316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,32,balanced,0.020197333147128422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,32,balanced,0.022367998957633972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,32,balanced,0.31285866101582843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,32,balanced,0.024186665813128155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,32,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,32,balanced,0.04099733382463455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,32,balanced,0.03572800010442734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,32,balanced,0.03470933437347412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,32,balanced,0.18339200814565024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,32,balanced,0.034314667185147606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,32,balanced,0.03590933233499527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,32,balanced,0.034330666065216064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,32,balanced,0.0344106654326121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,32,balanced,0.042170668641726174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,32,balanced,0.040720000863075256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,32,balanced,0.30792532364527386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,32,balanced,0.040336000422636666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,32,balanced,0.04658666749795278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,32,balanced,0.37407465775807697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,32,balanced,0.22857600450515747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,32,balanced,0.04849599798520406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,32,balanced,0.056927998860677086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,32,balanced,0.07163733243942261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,32,balanced,0.08790399630864461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,32,balanced,0.11006399989128113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,32,balanced,0.2723146677017212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,32,balanced,0.14669332901636759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,32,balanced,0.5672479867935181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,32,balanced,0.18772266308466592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,32,balanced,0.4090293248494466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,32,balanced,0.2650826573371887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,32,balanced,0.5721919933954874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,32,balanced,0.3424266576766968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,32,balanced,1.0895520051320393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,32,balanced,0.41633065541585285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,32,balanced,0.7657226721445719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,32,balanced,0.6623146533966064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,32,power_law_1.01,0.04229120016098022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,32,power_law_1.01,0.04084480106830597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,32,power_law_1.01,0.040275201201438904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,32,power_law_1.01,0.04300160109996796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,32,balanced,1.296015977859497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,32,power_law_1.01,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,32,power_law_1.01,0.04499199986457825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,32,power_law_1.01,0.04527359902858734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,32,power_law_1.01,0.046003198623657225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,32,power_law_1.01,0.04513919949531555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,32,power_law_1.01,0.046137601137161255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,32,power_law_1.01,0.04575360119342804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,32,power_law_1.01,0.0457152009010315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,32,power_law_1.01,0.04702079892158508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,32,power_law_1.01,0.04723840057849884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,32,power_law_1.01,0.04976640045642853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,32,power_law_1.01,0.05018879771232605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,32,power_law_1.01,0.051660799980163576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,32,power_law_1.01,0.05736960172653198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,32,power_law_1.01,0.06172800064086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,32,power_law_1.01,0.06839680075645446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,32,power_law_1.01,0.07587839961051941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,32,power_law_1.01,0.10156799554824829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,32,power_law_1.01,0.11678080558776856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,32,power_law_1.01,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,32,power_law_1.01,0.14545279741287231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,32,power_law_1.01,0.037574398517608645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,32,power_law_1.01,0.17608319520950316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,32,power_law_1.01,0.03660799860954285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,32,power_law_1.01,0.27445759773254397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,32,power_law_1.01,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,32,power_law_1.01,0.3207103967666626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,32,power_law_1.01,0.0412992000579834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,32,power_law_1.01,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,32,power_law_1.01,0.4056447982788086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,32,power_law_1.01,0.0390720009803772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,32,power_law_1.01,0.04390400052070618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,32,power_law_1.01,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,32,power_law_1.01,0.6470079898834229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,32,power_law_1.01,0.042028799653053284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,32,power_law_1.01,0.04099839925765991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,32,power_law_1.01,1.2484736442565918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,32,power_law_1.01,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,32,power_law_1.01,0.04208639860153198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,32,power_law_1.01,0.042822399735450746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,32,power_law_1.01,0.04384639859199524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,32,power_law_1.01,0.043680000305175784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,32,power_law_1.01,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,32,power_law_1.01,0.04442879855632782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,32,power_law_1.01,0.04490239918231964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,32,power_law_1.01,0.04421760141849518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,32,power_law_1.01,0.04497919976711273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,32,power_law_1.01,0.044684800505638125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,32,power_law_1.01,0.045465600490570066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,32,power_law_1.01,0.045407998561859134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,32,power_law_1.01,0.04547199904918671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,32,power_law_1.01,0.04843519926071167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,32,power_law_1.01,0.046419200301170346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,32,power_law_1.01,0.049004799127578734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,32,power_law_1.01,0.04656639993190766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,32,power_law_1.01,0.05262079834938049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,32,power_law_1.01,0.046323201060295104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,32,power_law_1.01,0.05648639798164368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,32,power_law_1.01,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,32,power_law_1.01,0.05124480128288269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,32,power_law_1.01,0.06107519865036011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,32,power_law_1.01,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,32,power_law_1.01,0.07156479954719544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,32,power_law_1.01,0.05801600217819214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,32,power_law_1.01,0.09071999788284302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,32,power_law_1.01,0.11454720497131347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.01,0.06173440217971802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.01,0.07084800004959106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,32,power_law_1.01,0.1530751943588257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.01,0.08457599878311158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,32,power_law_1.01,0.19271680116653442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.01,0.09998720288276672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,32,power_law_1.01,0.2304447889328003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.01,0.11969920396804809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,32,power_law_1.01,0.33904640674591063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.01,0.16172159910202027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,32,power_law_1.01,0.46151041984558105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.01,0.19566080570220948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,32,power_law_1.01,0.5876224040985107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.01,0.27578880786895754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,32,power_law_1.01,0.8984895706176758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.01,0.3701375961303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,32,power_law_1.01,1.818387222290039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,32,power_law_1.01,0.4609344005584717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,32,power_law_1.01,0.7323071956634521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,32,power_law_1.01,1.518175983428955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,32,power_law_1.01,0.017817600071430205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,32,power_law_1.01,0.018592000007629395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,32,power_law_1.01,0.01828480064868927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,32,power_law_1.01,0.01897599995136261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,32,power_law_1.01,0.020467199385166168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,32,power_law_1.01,0.022873599827289582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,32,power_law_1.01,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,32,power_law_1.01,0.03994880020618439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,32,power_law_1.01,0.03386879861354828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,32,power_law_1.01,0.033062401413917544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,32,power_law_1.01,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,32,power_law_1.01,0.033964800834655764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,32,power_law_1.01,0.033632001280784606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,32,power_law_1.01,0.032620799541473386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,32,power_law_1.01,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,32,power_law_1.01,0.03489919900894165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,32,power_law_1.01,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,32,power_law_1.01,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.01,0.04337919950485229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.01,0.05043839812278748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.01,0.055340802669525145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.01,0.06855679750442505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.01,0.08672639727592468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.01,0.11437439918518066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.01,0.1450495958328247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.01,0.19409279823303222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.01,0.25747199058532716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,32,power_law_1.01,0.30858240127563474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,32,power_law_1.01,0.4867839813232422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,32,power_law_1.01,0.9461055755615234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,32,power_law_1.2,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,32,power_law_1.2,0.03772160112857818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,32,power_law_1.2,0.03813759982585907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,32,power_law_1.2,0.04097920060157776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,32,power_law_1.2,0.042931199073791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,32,power_law_1.2,0.04439679980278015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,32,power_law_1.2,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,32,power_law_1.2,0.04673280119895935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,32,power_law_1.2,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,32,power_law_1.2,0.045158401131629944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,32,power_law_1.2,0.04500479996204376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,32,power_law_1.2,0.045772799849510194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,32,power_law_1.2,0.04575999975204468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,32,power_law_1.2,0.04749439954757691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,32,power_law_1.2,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,32,power_law_1.2,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,32,power_law_1.2,0.05369600057601929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,32,power_law_1.2,0.057708799839019775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,32,power_law_1.2,0.0627135992050171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,32,power_law_1.2,0.0754751980304718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,32,power_law_1.2,0.0861631989479065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,32,power_law_1.2,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,32,power_law_1.2,0.10957440137863159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,32,power_law_1.2,0.03685120046138764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,32,power_law_1.2,0.1313024044036865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,32,power_law_1.2,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,32,power_law_1.2,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,32,power_law_1.2,0.19179519414901733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,32,power_law_1.2,0.04281600117683411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,32,power_law_1.2,0.0379584014415741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,32,power_law_1.2,0.2640255928039551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,32,power_law_1.2,0.03999359905719757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,32,power_law_1.2,0.03802880048751831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,32,power_law_1.2,0.3936768054962158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,32,power_law_1.2,0.04178560078144074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,32,power_law_1.2,0.03925760090351105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,32,power_law_1.2,0.5258880138397217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,32,power_law_1.2,0.04102399945259094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,32,power_law_1.2,0.040940800309181215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,32,power_law_1.2,0.04215039908885956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,32,power_law_1.2,0.6358911991119385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,32,power_law_1.2,0.043507200479507444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,32,power_law_1.2,0.0418368011713028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,32,power_law_1.2,0.04439679980278015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,32,power_law_1.2,0.7815423965454101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,32,power_law_1.2,0.04348799884319306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,32,power_law_1.2,0.044326400756835936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,32,power_law_1.2,2.1620864868164062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,32,power_law_1.2,0.04239999949932098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,32,power_law_1.2,0.045075199007987975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,32,power_law_1.2,0.0445248007774353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,32,power_law_1.2,0.045337599515914914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,32,power_law_1.2,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,32,power_law_1.2,0.0444927990436554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,32,power_law_1.2,0.046316799521446225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,32,power_law_1.2,0.04555520117282867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,32,power_law_1.2,0.047225600481033324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,32,power_law_1.2,0.04938240051269531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,32,power_law_1.2,0.04743039906024933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,32,power_law_1.2,0.050271999835968015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,32,power_law_1.2,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,32,power_law_1.2,0.05374720096588135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,32,power_law_1.2,0.05232639908790589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,32,power_law_1.2,0.059308797121047974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,32,power_law_1.2,0.055315202474594115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,32,power_law_1.2,0.06653439998626709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,32,power_law_1.2,0.06030719876289368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,32,power_law_1.2,0.0759872019290924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.2,0.06749439835548401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,32,power_law_1.2,0.09336959719657897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.2,0.07969279885292054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,32,power_law_1.2,0.13088639974594116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.2,0.08869119882583618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,32,power_law_1.2,0.15584640502929686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.2,0.11390080451965331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,32,power_law_1.2,0.23758080005645751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.2,0.12775039672851562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,32,power_law_1.2,0.2869951963424683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.2,0.20234239101409912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,32,power_law_1.2,0.4477952003479004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.2,0.25504000186920167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,32,power_law_1.2,0.5704192161560059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.2,0.41123199462890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,32,power_law_1.2,0.7868607997894287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.2,0.4011392116546631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,32,power_law_1.2,1.1911104202270508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,32,power_law_1.2,0.683897590637207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,32,power_law_1.2,2.5192447662353517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,32,power_law_1.2,1.2680447578430176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,32,power_law_1.2,1.9291584014892578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,32,power_law_1.2,0.018566399812698364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,32,power_law_1.2,0.01865600049495697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,32,power_law_1.2,0.019020800292491914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,32,power_law_1.2,0.019174399971961974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,32,power_law_1.2,0.021158400177955627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,32,power_law_1.2,0.023494400084018707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,32,power_law_1.2,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,32,power_law_1.2,0.039776000380516055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,32,power_law_1.2,0.033792001008987424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,32,power_law_1.2,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,32,power_law_1.2,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,32,power_law_1.2,0.03371520042419433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,32,power_law_1.2,0.03290880024433136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,32,power_law_1.2,0.032767999172210696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,32,power_law_1.2,0.03346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,32,power_law_1.2,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,32,power_law_1.2,0.03769600093364715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,32,power_law_1.2,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.2,0.04439040124416351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.2,0.05008640289306641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.2,0.055820798873901366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.2,0.07937920093536377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.2,0.08889600038528442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.2,0.11871360540390015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.2,0.14701440334320068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.2,0.21434879302978516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.2,0.28834559917449953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,32,power_law_1.2,0.3404736042022705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,32,power_law_1.2,0.5439871788024903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,32,power_law_1.2,1.1069120407104491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,16,balanced,0.03654933224121729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,16,balanced,0.03819733361403147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,16,balanced,0.039173332353432976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,16,balanced,0.04284800092379252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,16,balanced,0.047210668524106346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,16,balanced,0.04720533390839895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,16,balanced,0.046623999873797096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,16,balanced,0.0466186652580897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,16,balanced,0.04755199948946635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,16,balanced,0.04675200084845225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,16,balanced,0.04655999938646952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,16,balanced,0.04872000217437744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,16,balanced,0.047055999437967934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,16,balanced,0.04863466819127401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,16,balanced,0.05340800185998281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,16,balanced,0.050848002235094704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,16,balanced,0.054378668467203774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,16,balanced,0.05438933273156484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,16,balanced,0.05725333094596863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,16,balanced,0.06063466767470042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,16,balanced,0.06651733318964641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,16,power_law_1.01,0.04590719938278198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,16,power_law_1.01,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,16,power_law_1.01,0.03898240029811859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,16,power_law_1.01,0.042630401253700254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,16,balanced,0.08231466511885326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,16,balanced,0.035930665830771126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,16,power_law_1.01,0.043961599469184875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,16,balanced,0.036474667489528656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,16,power_law_1.01,0.04695680141448975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,16,balanced,0.03732266773780187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,16,power_law_1.01,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,16,balanced,0.04092800120512644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,16,power_law_1.01,0.04643200039863586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,16,balanced,0.04271999994913737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,16,balanced,0.09130133191744487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,16,balanced,0.045312002301216125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,16,power_law_1.01,0.046700799465179445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,16,balanced,0.04654400050640106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,16,power_law_1.01,0.04721280038356781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,16,balanced,0.04436799883842468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,16,power_law_1.01,0.046675199270248414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,16,balanced,0.04527466495831808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,16,power_law_1.01,0.04731520116329193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,16,balanced,0.04679466784000397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,16,balanced,0.11343466242154439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,16,power_law_1.01,0.047577598690986635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,16,balanced,0.04710933566093445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,16,power_law_1.01,0.04902400076389313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,16,balanced,0.04913066824277242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,16,power_law_1.01,0.05199360251426697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,16,balanced,0.0468746672074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,16,power_law_1.01,0.051948797702789304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,16,balanced,0.04808533191680908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,16,power_law_1.01,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,16,balanced,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,16,balanced,0.12981333335240683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,16,power_law_1.01,0.061247998476028444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,16,balanced,0.05243200063705444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,16,power_law_1.01,0.06423680186271667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,16,balanced,0.05232533315817515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,16,power_law_1.01,0.07411199808120728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,16,balanced,0.056890666484832764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,16,power_law_1.01,0.08511999845504761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,16,power_law_1.01,0.10220160484313964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,16,balanced,0.06055466830730438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,16,power_law_1.01,0.12254079580307006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,16,power_law_1.01,0.1571903944015503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,16,balanced,0.06587199866771698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,16,power_law_1.01,0.2032320022583008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,16,balanced,0.17045332988103232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,16,power_law_1.01,0.27489919662475587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,16,balanced,0.07513066629568736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,16,power_law_1.01,0.3793728113174438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,16,power_law_1.01,0.4251711845397949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,16,balanced,0.1033066709836324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,16,power_law_1.01,0.682528018951416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,16,power_law_1.01,1.3772735595703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,16,balanced,0.12730133533477783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,16,balanced,0.22014933824539185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,16,balanced,0.1644053359826406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,16,balanced,0.20204800367355347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,16,balanced,0.2590240041414897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,16,balanced,0.27585599819819134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,16,power_law_1.01,0.04420480132102966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,16,balanced,0.35386133193969727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,16,power_law_1.01,0.03847039937973022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,16,power_law_1.01,0.0367935985326767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,16,power_law_1.01,0.0401856005191803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,16,power_law_1.01,0.04121600091457367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,16,power_law_1.01,0.04349440038204193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,16,power_law_1.01,0.04339199960231781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,16,power_law_1.01,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,16,balanced,0.3813386758168538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,16,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,16,power_law_1.01,0.04538240134716034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,16,power_law_1.01,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,16,power_law_1.01,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,16,power_law_1.01,0.0473471999168396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,16,power_law_1.01,0.04745599925518036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,16,balanced,0.4296106497446696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,16,power_law_1.01,0.05230720043182373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,16,power_law_1.01,0.05306879878044128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,16,power_law_1.01,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,16,power_law_1.01,0.06059520244598389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,16,power_law_1.01,0.0667967975139618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,16,power_law_1.01,0.08028159737586975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,16,power_law_1.01,0.09521920084953309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,16,power_law_1.01,0.12447359561920165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,16,power_law_1.01,0.1471935987472534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,16,power_law_1.01,0.21273601055145264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,16,power_law_1.01,0.2557055950164795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,16,power_law_1.01,0.3779583930969238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,16,power_law_1.01,0.48570880889892576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,16,power_law_1.01,0.6095871925354004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,16,power_law_1.01,0.915833568572998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,16,power_law_1.01,2.0334848403930663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,16,balanced,0.6580640077590942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,16,balanced,0.7189493179321289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,16,power_law_1.2,0.04663040041923523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,16,power_law_1.2,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,16,power_law_1.2,0.03924480080604553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,16,power_law_1.2,0.04296320080757141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,16,power_law_1.2,0.04385280013084412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,16,power_law_1.2,0.04513919949531555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,16,power_law_1.2,0.04672000110149384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,16,power_law_1.2,0.0462336003780365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,16,power_law_1.2,0.04682239890098572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,16,power_law_1.2,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,16,power_law_1.2,0.047731199860572816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,16,power_law_1.2,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,16,power_law_1.2,0.04775680005550385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,16,power_law_1.2,0.04875519871711731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,16,power_law_1.2,0.052748799324035645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,16,power_law_1.2,0.05183359980583191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,16,balanced,1.278981367746989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,16,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,16,power_law_1.2,0.06058880090713501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,16,power_law_1.2,0.06857600212097167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,16,power_law_1.2,0.07654399871826172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,16,power_law_1.2,0.08982399702072144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,16,power_law_1.2,0.10426880121231079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,16,power_law_1.2,0.12918399572372435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,16,power_law_1.2,0.18595199584960936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,16,power_law_1.2,0.24579839706420897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,16,power_law_1.2,0.36003201007843016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,16,power_law_1.2,0.4141823768615723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,16,power_law_1.2,0.6741119861602783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,16,power_law_1.2,1.133407974243164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,16,power_law_1.2,2.1319679260253905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,16,power_law_1.2,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,16,power_law_1.2,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,16,power_law_1.2,0.03927040100097656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,16,power_law_1.2,0.04043520092964172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,16,power_law_1.2,0.041766399145126344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,16,power_law_1.2,0.042771199345588685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,16,power_law_1.2,0.04392960071563721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,16,power_law_1.2,0.044300800561904906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,16,power_law_1.2,0.043884798884391785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,16,power_law_1.2,0.045542401075363156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,16,power_law_1.2,0.04636160135269165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,16,power_law_1.2,0.04665600061416626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,16,power_law_1.2,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,16,power_law_1.2,0.04695680141448975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,16,power_law_1.2,0.05191680192947388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,16,power_law_1.2,0.052985602617263795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,16,power_law_1.2,0.055219197273254396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,8,balanced,0.03643200049797694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,16,power_law_1.2,0.061247998476028444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,8,balanced,0.03919466584920883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,16,power_law_1.2,0.07059199810028076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,8,balanced,0.0408746674656868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,8,balanced,0.044719999035199486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,16,power_law_1.2,0.09203839898109437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,8,balanced,0.04854399959246317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,16,power_law_1.2,0.10017919540405273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,8,balanced,0.05162666738033295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,16,power_law_1.2,0.13630080223083496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,8,balanced,0.05352533360322317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,16,power_law_1.2,0.16554239988327027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,8,balanced,0.05121066669623057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,8,balanced,0.05030933519204458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,16,power_law_1.2,0.23539199829101562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,8,balanced,0.05186666548252106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,16,power_law_1.2,0.29153919219970703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,8,balanced,0.05137600004673004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,16,power_law_1.2,0.4394688129425049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,8,balanced,0.05136533578236898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,16,power_law_1.2,0.6411712169647217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,8,balanced,0.0524533341328303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,16,power_law_1.2,0.8464320182800293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,8,balanced,0.052602668603261314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,16,power_law_1.2,1.3949376106262208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,8,balanced,0.054976001381874084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,8,balanced,0.05678399900595347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,16,power_law_1.2,2.4957632064819335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,8,balanced,0.05691733459631602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,8,balanced,0.06460799773534139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,8,balanced,0.06620799998442332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,8,balanced,0.07221866647402446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,8,balanced,0.07996800045172374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,8,balanced,0.0345920001467069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,8,balanced,0.03676266719897588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,8,balanced,0.038634667793909706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,8,balanced,0.042768001556396484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,8,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,8,balanced,0.09991466999053955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,8,balanced,0.048512001832326256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,8,balanced,0.050623998045921326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,8,balanced,0.048901334404945374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,8,balanced,0.04862933357556661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,8,balanced,0.0505973349014918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,8,balanced,0.11163199941317241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,8,balanced,0.050928001602490745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,8,balanced,0.05067199965318044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,8,balanced,0.050517335534095764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,8,balanced,0.0510506679614385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,8,balanced,0.058634668588638306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,8,balanced,0.14146133263905844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,8,balanced,0.056618665655454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,8,balanced,0.057573333382606506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,8,power_law_1.01,0.04595839977264404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,8,power_law_1.01,0.04204800128936768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,8,balanced,0.06251200040181477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,8,power_law_1.01,0.039654400944709775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,8,balanced,0.06906666855017345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,8,power_law_1.01,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,8,balanced,0.17339199781417847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,8,power_law_1.01,0.04647679924964905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,8,balanced,0.0762613316377004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,8,power_law_1.01,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,8,power_law_1.01,0.048825600743293764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,8,power_law_1.01,0.04880000054836273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,8,balanced,0.09103999535242717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,8,power_law_1.01,0.05001599788665771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,8,power_law_1.01,0.05014399886131286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,8,power_law_1.01,0.05077760219573975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,8,balanced,0.12470933794975281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,8,power_law_1.01,0.050323200225830075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,8,balanced,0.24318933486938477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,8,power_law_1.01,0.05263360142707825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,8,power_law_1.01,0.05310720205307007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,8,power_law_1.01,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,8,balanced,0.1462506651878357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,8,power_law_1.01,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,8,power_law_1.01,0.061324799060821535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,8,power_law_1.01,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,8,power_law_1.01,0.0753600001335144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,8,power_law_1.01,0.08630399703979492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,8,balanced,0.19751467307408652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,8,power_law_1.01,0.09804159998893738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,8,power_law_1.01,0.11282559633255004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,8,power_law_1.01,0.14016000032424927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,8,balanced,0.3012106617291768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,8,power_law_1.01,0.19706239700317382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,8,power_law_1.01,0.2499135971069336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,8,power_law_1.01,0.31366400718688964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,8,balanced,0.24489599466323853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,8,power_law_1.01,0.4633344173431396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,8,power_law_1.01,0.531660795211792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,8,power_law_1.01,0.8293631553649903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,8,power_law_1.01,1.7200960159301757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,8,balanced,0.34621334075927734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,8,balanced,0.3734240134557088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,8,balanced,0.44202133019765216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,8,balanced,0.5487786531448364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,8,balanced,0.5598400036493937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,8,power_law_1.01,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,8,power_law_1.01,0.03909119963645935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,8,power_law_1.01,0.03654400110244751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,8,power_law_1.01,0.04174720048904419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,8,power_law_1.01,0.0444927990436554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,8,power_law_1.01,0.04590719938278198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,8,power_law_1.01,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,8,power_law_1.01,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,8,power_law_1.01,0.048492801189422605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,8,power_law_1.01,0.04848000109195709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,8,power_law_1.01,0.050387197732925416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,8,power_law_1.01,0.04906240105628967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,8,power_law_1.01,0.049184000492095946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,8,power_law_1.01,0.05262719988822937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,8,power_law_1.01,0.056377601623535153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,8,balanced,0.8448586463928223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,8,power_law_1.01,0.05845119953155518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,8,power_law_1.01,0.06017919778823853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,8,power_law_1.01,0.06687359809875489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,8,power_law_1.01,0.07533439993858337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,8,power_law_1.01,0.09390079975128174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,8,power_law_1.01,0.10930559635162354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,8,power_law_1.01,0.1509760022163391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,8,power_law_1.01,0.1765120029449463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,8,power_law_1.01,0.23418879508972168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,8,power_law_1.01,0.313811206817627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,8,power_law_1.01,0.46737279891967776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,8,power_law_1.01,0.5575168132781982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,8,power_law_1.01,0.6934016227722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,8,power_law_1.01,1.2047360420227051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,8,power_law_1.01,2.496019172668457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,8,balanced,1.0717493693033855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,8,balanced,1.6500213940938313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,8,power_law_1.2,0.045900800824165346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,8,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,8,power_law_1.2,0.040678399801254275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,8,power_law_1.2,0.04334079921245575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,8,power_law_1.2,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,8,power_law_1.2,0.0489984005689621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,8,power_law_1.2,0.04927360117435455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,8,power_law_1.2,0.05041279792785645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,8,power_law_1.2,0.04984320104122162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,8,power_law_1.2,0.05062400102615357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,8,power_law_1.2,0.05070080161094666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,8,power_law_1.2,0.05297279953956604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,8,power_law_1.2,0.052716797590255736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,8,power_law_1.2,0.054502397775650024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,8,power_law_1.2,0.05675519704818725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,8,power_law_1.2,0.05941759943962097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,8,power_law_1.2,0.06222079992294312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,8,power_law_1.2,0.0720192015171051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,8,power_law_1.2,0.0773311972618103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,8,power_law_1.2,0.09164159893989562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,8,power_law_1.2,0.10289920568466186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,8,power_law_1.2,0.1257599949836731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,8,power_law_1.2,0.1655743956565857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,8,power_law_1.2,0.22785279750823975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,8,power_law_1.2,0.2947648048400879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,8,power_law_1.2,0.3934463977813721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,8,power_law_1.2,0.5989503860473633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,8,power_law_1.2,0.8000191688537598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,8,power_law_1.2,1.252569580078125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,8,power_law_1.2,2.513088035583496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,8,power_law_1.2,0.04230400025844574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,8,power_law_1.2,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,8,power_law_1.2,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,4,balanced,0.03811733424663544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,8,power_law_1.2,0.04088320136070252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,4,balanced,0.04022400081157684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,4,balanced,0.04279999931653341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,8,power_law_1.2,0.04245760142803192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,4,balanced,0.04870399832725525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,8,power_law_1.2,0.046758401393890384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,4,balanced,0.055349335074424744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,8,power_law_1.2,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,4,balanced,0.058133333921432495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,8,power_law_1.2,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,4,balanced,0.05932266513506571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,4,balanced,0.05740800003210703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,8,power_law_1.2,0.048640000820159915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,4,balanced,0.058543999989827476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,8,power_law_1.2,0.04890879988670349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,4,balanced,0.05907199780146281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,8,power_law_1.2,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,4,balanced,0.05705599983533224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,8,power_law_1.2,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,4,balanced,0.058559998869895935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,8,power_law_1.2,0.051801598072052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,4,balanced,0.05914133290449778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,8,power_law_1.2,0.053830397129058835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,4,balanced,0.06065066655476888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,8,power_law_1.2,0.05877760052680969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,4,balanced,0.06551466882228851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,8,power_law_1.2,0.05869439840316772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,8,power_law_1.2,0.06357120275497437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,4,balanced,0.06367466847101848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,8,power_law_1.2,0.06901119947433472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,4,balanced,0.06745600203673045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,8,power_law_1.2,0.07782400250434876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,4,balanced,0.07247999807198842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,8,power_law_1.2,0.09753599762916565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,8,power_law_1.2,0.115065598487854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,4,balanced,0.07948799928029378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,8,power_law_1.2,0.15820159912109374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,8,power_law_1.2,0.20017919540405274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,4,balanced,0.09129599730173747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,8,power_law_1.2,0.27533440589904784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,8,power_law_1.2,0.3387648105621338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,4,balanced,0.10145066181818645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,8,power_law_1.2,0.4715583801269531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,4,balanced,0.03738133360942205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,8,power_law_1.2,0.646131181716919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,4,balanced,0.039488000174363456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,4,balanced,0.12972799936930338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,8,power_law_1.2,0.8579520225524903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,4,balanced,0.041135999063650765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,4,balanced,0.046816001335779824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,8,power_law_1.2,1.5011520385742188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,4,balanced,0.054832001527150474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,4,balanced,0.05695466697216034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,4,balanced,0.1474720040957133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,8,power_law_1.2,3.3845375061035154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,4,balanced,0.05958933134873708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,4,balanced,0.05712533493836721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,4,balanced,0.05691733459631602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,4,balanced,0.057706668972969055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,4,balanced,0.20403200387954712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,4,balanced,0.05929600199063619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,4,balanced,0.061477333307266235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,4,balanced,0.06003733476003011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,4,balanced,0.06115733087062836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,4,balanced,0.06714666883150737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,4,balanced,0.0670773337284724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,4,balanced,0.25177599986394245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,4,balanced,0.06941333413124084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,4,balanced,0.07692266503969829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,4,balanced,0.08341333270072937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,4,balanced,0.09992532928784688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,4,balanced,0.3657706578572591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,4,balanced,0.12153066198031108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,4,balanced,0.16133333245913187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,4,balanced,0.19139200448989868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,4,power_law_1.01,0.04441600143909454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,4,balanced,0.46508800983428955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,4,power_law_1.01,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,4,power_law_1.01,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,4,balanced,0.2638079921404521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,4,power_law_1.01,0.04625279903411865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,4,power_law_1.01,0.049491199851036075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,4,power_law_1.01,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,4,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,4,power_law_1.01,0.05605120062828064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,4,power_law_1.01,0.05656319856643677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,4,power_law_1.01,0.05751680135726929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,4,balanced,0.336677352587382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,4,power_law_1.01,0.057036799192428586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,4,power_law_1.01,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,4,power_law_1.01,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,4,balanced,0.5718186696370443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,4,power_law_1.01,0.06017919778823853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,4,power_law_1.01,0.06592000126838685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,4,power_law_1.01,0.06822400093078614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,4,power_law_1.01,0.0720255970954895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,4,power_law_1.01,0.08179839849472045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,4,power_law_1.01,0.08966400027275086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,4,power_law_1.01,0.1016767978668213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,4,balanced,0.4914666811625163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,4,power_law_1.01,0.11963520050048829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,4,power_law_1.01,0.15582079887390138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,4,power_law_1.01,0.18407039642333983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,4,power_law_1.01,0.2731647968292236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,4,power_law_1.01,0.31592960357666017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,4,power_law_1.01,0.446995210647583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,4,power_law_1.01,0.6018176078796387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,4,power_law_1.01,0.714739179611206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,4,power_law_1.01,1.099948787689209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,4,power_law_1.01,2.3430208206176757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,4,balanced,0.86626664797465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,4,balanced,0.6329226493835449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,4,balanced,0.787887970606486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,4,power_law_1.01,0.042131200432777405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,4,power_law_1.01,0.040243199467658995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,4,power_law_1.01,0.0403328001499176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,4,power_law_1.01,0.0434112012386322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,4,power_law_1.01,0.046982398629188536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,4,power_law_1.01,0.054073601961135864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,4,power_law_1.01,0.054579198360443115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,4,power_law_1.01,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,4,power_law_1.01,0.05495679974555969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,4,balanced,1.6678719520568848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,4,balanced,1.22651203473409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,4,power_law_1.01,0.05612800121307373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,4,power_law_1.01,0.058259201049804685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,4,power_law_1.01,0.05985280275344849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,4,power_law_1.01,0.059539198875427246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,4,power_law_1.01,0.06146559715270996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,4,power_law_1.01,0.06606720089912414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,4,power_law_1.01,0.06712960004806519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,4,power_law_1.01,0.07199360132217407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,4,power_law_1.01,0.08141440153121948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,4,power_law_1.01,0.09423999786376953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,4,power_law_1.01,0.12200959920883178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,4,power_law_1.01,0.13691519498825072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,4,power_law_1.01,0.17828480005264283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,4,power_law_1.01,0.22724480628967286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,4,power_law_1.01,0.32598400115966797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,4,power_law_1.01,0.388319993019104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,4,power_law_1.01,0.575219202041626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,4,power_law_1.01,0.7229568004608155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,4,power_law_1.01,0.9387904167175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,4,power_law_1.01,1.6531328201293944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,4,power_law_1.01,2.94256649017334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,4,balanced,2.387808005015055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,4,power_law_1.2,0.04454399943351746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,4,power_law_1.2,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,4,power_law_1.2,0.04234879910945892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,4,power_law_1.2,0.045363199710845944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,4,power_law_1.2,0.04912000000476837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,4,power_law_1.2,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,4,power_law_1.2,0.054502397775650024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,4,power_law_1.2,0.05631999969482422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,4,power_law_1.2,0.05639680027961731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,4,power_law_1.2,0.056569600105285646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,4,power_law_1.2,0.05856639742851257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,4,power_law_1.2,0.05936639904975891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,4,power_law_1.2,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,4,power_law_1.2,0.06216319799423218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,4,power_law_1.2,0.06654080152511596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,4,power_law_1.2,0.07041280269622803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,4,power_law_1.2,0.07375360131263733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,4,power_law_1.2,0.0838271975517273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,4,power_law_1.2,0.09341440200805665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,4,power_law_1.2,0.10651520490646363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,4,power_law_1.2,0.11910400390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,4,power_law_1.2,0.1464192032814026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,4,power_law_1.2,0.18869760036468505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,4,power_law_1.2,0.2803647994995117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,4,power_law_1.2,0.3603264093399048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,4,power_law_1.2,0.5419392108917236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,4,power_law_1.2,0.6899263858795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,4,power_law_1.2,0.9514240264892578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,4,power_law_1.2,1.3936063766479492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,4,power_law_1.2,2.770060729980469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,4,power_law_1.2,0.042156800627708435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,4,power_law_1.2,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,4,power_law_1.2,0.04017280042171478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,4,power_law_1.2,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,4,power_law_1.2,0.04694400131702423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,4,power_law_1.2,0.05375999808311462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,4,power_law_1.2,0.05490559935569763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,4,power_law_1.2,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,4,power_law_1.2,0.05680000185966492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,4,power_law_1.2,0.05616639852523804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,4,power_law_1.2,0.05731199979782105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,4,power_law_1.2,0.05907840132713318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,4,power_law_1.2,0.060134398937225345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,4,power_law_1.2,0.06117759943008423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,4,power_law_1.2,0.06686720252037048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,4,power_law_1.2,0.06883199810981751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,4,power_law_1.2,0.07432960271835327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,4,power_law_1.2,0.0846783995628357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,4,power_law_1.2,0.09562240242958069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,32,balanced,0.05671466886997223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,4,power_law_1.2,0.12432639598846436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,32,balanced,0.05765866736570994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,4,power_law_1.2,0.1411072015762329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,32,balanced,0.05705599983533224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,32,balanced,0.06484266618887584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,4,power_law_1.2,0.19371520280838012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,32,balanced,0.0720000018676122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,4,power_law_1.2,0.2425407886505127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,32,balanced,0.08284266789754231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,4,power_law_1.2,0.3295488119125366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,32,balanced,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,4,power_law_1.2,0.4285439968109131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,32,balanced,0.0817386656999588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,32,balanced,0.08253333469231923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,4,power_law_1.2,0.6139840126037598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,32,balanced,0.08337066570917766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,4,power_law_1.2,0.9041472434997558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,32,balanced,0.08320533235867818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,4,power_law_1.2,1.040454387664795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,32,balanced,0.08365866541862488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,32,balanced,0.08176533381144206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,4,power_law_1.2,1.7250879287719727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,32,balanced,0.08507733543713887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,4,power_law_1.2,3.0738752365112303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,32,balanced,0.08762666583061218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,32,balanced,0.09008000294367473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,32,balanced,0.0925546685854594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,32,balanced,0.09718400239944458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,32,balanced,0.10071999828020732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,32,balanced,0.05038933455944061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,32,balanced,0.10878933469454448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,32,balanced,0.0505920002857844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,32,balanced,0.052895997961362205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,32,balanced,0.057573333382606506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,32,balanced,0.11749333143234253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,32,balanced,0.06100266675154368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,32,balanced,0.06509333352247874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,32,balanced,0.0639519989490509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,32,balanced,0.0678773323694865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,32,balanced,0.14079999923706055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,32,balanced,0.06910400092601776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,32,balanced,0.0659093310435613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,32,balanced,0.06966400146484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,32,balanced,0.06716266771157582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,32,balanced,0.15498133500417074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,32,balanced,0.07127999762694041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,32,balanced,0.07111999889214833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,32,balanced,0.07364800075689952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,32,balanced,0.07502399881680806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,32,balanced,0.07805866499741872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,32,balanced,0.19598400592803955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,32,balanced,0.08539733290672302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,32,balanced,0.09149332841237386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,32,balanced,0.11385066310564677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,32,balanced,0.1297546625137329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,32,balanced,0.24424533049265543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,32,balanced,0.1651040017604828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,32,balanced,0.1929759979248047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,32,balanced,0.3076159954071045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,32,balanced,0.25545599063237506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,32,balanced,0.05063466727733612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,32,balanced,0.052853330969810486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,32,balanced,0.05274133384227753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,32,balanced,0.06286400059858958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,32,balanced,0.06308266520500183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,32,balanced,0.06946666538715363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,32,balanced,0.06743999818960826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,32,balanced,0.32313066720962524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,32,balanced,0.07149866720040639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,32,balanced,0.06938133140405019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,32,balanced,0.06904000043869019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,32,balanced,0.06852266689141591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,32,balanced,0.40223467350006104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,32,balanced,0.07180266578992207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,32,balanced,0.07213866710662842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,32,balanced,0.07148799796899159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,32,balanced,0.07744533320267995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,32,balanced,0.07436800003051758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,32,balanced,0.43827199935913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,32,balanced,0.07935466865698497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,32,balanced,0.08354666829109192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,32,balanced,0.08933867017428081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,32,balanced,0.09810133775075276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,32,balanced,0.10946666200955708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,32,balanced,0.4585119883219401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,32,balanced,0.5683146715164185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,32,balanced,0.14415466785430908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,32,balanced,0.1614773372809092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,32,balanced,0.20787199338277182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,32,balanced,0.6813066800435384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,32,balanced,0.2566399971644084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,32,balanced,0.7039840221405029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,32,balanced,0.3439306815465291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,32,balanced,0.023562667270501454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,32,balanced,0.02458133300145467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,32,balanced,0.024671999116738636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,32,balanced,0.024885334074497223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,32,balanced,0.026373334228992462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,32,balanced,0.4418826500574748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,32,balanced,1.0528480211893718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,32,balanced,0.03013866643110911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,32,balanced,0.051829333106676735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,32,balanced,0.050848002235094704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,32,balanced,0.05119466781616211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,32,balanced,0.049039999643961586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,32,balanced,0.04866666595141093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,32,balanced,0.04713066418965658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,32,balanced,0.04527466495831808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,32,balanced,0.04452266792456309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,32,balanced,0.054570664962132774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,32,balanced,0.05715733269850413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,32,balanced,0.049173335234324135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,32,balanced,0.5268959999084473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,32,balanced,0.05559466779232025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,32,balanced,0.07559466858704884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,32,balanced,0.08681600292523702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,32,balanced,0.1083573301633199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,32,balanced,0.1402239998181661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,32,balanced,0.17341333627700806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,32,balanced,1.3592212994893391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,32,balanced,0.798362652460734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,32,balanced,0.24079465866088867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,32,balanced,0.3049013415972392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,32,balanced,2.0670560201009116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,32,balanced,0.4426079988479614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,32,balanced,0.5863733291625977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,32,balanced,1.5631732940673828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,32,balanced,0.7262880007425944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,32,balanced,1.1428106625874836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,32,power_law_1.01,0.08039680123329163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,32,power_law_1.01,0.08023040294647217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,32,power_law_1.01,0.06572160124778748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,32,power_law_1.01,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,32,power_law_1.01,0.07422720193862915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,32,power_law_1.01,0.0798143982887268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,32,power_law_1.01,0.07991679906845092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,32,power_law_1.01,0.08252800107002259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,32,power_law_1.01,0.08092799782752991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,32,power_law_1.01,0.07957119941711426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,32,power_law_1.01,0.0804095983505249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,32,power_law_1.01,0.08199679851531982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,32,power_law_1.01,0.08210560083389282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,32,power_law_1.01,0.08376960158348083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,32,power_law_1.01,0.088128000497818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,32,power_law_1.01,0.08761600255966187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,32,power_law_1.01,0.09359359741210938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,32,power_law_1.01,0.10132479667663574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,32,power_law_1.01,0.06345599889755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,32,power_law_1.01,0.10808320045471191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,32,balanced,2.246789296468099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,32,power_law_1.01,0.05801600217819214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,32,power_law_1.01,0.12400000095367432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,32,power_law_1.01,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,32,power_law_1.01,0.144595205783844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,32,power_law_1.01,0.0585536003112793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,32,power_law_1.01,0.17878400087356566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,32,power_law_1.01,0.061459201574325564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,32,power_law_1.01,0.20624639987945556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,32,power_law_1.01,0.06327040195465088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,32,power_law_1.01,0.2848320007324219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,32,power_law_1.01,0.06503040194511414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,32,power_law_1.01,0.34666240215301514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,32,power_law_1.01,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,32,power_law_1.01,0.47672319412231445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,32,power_law_1.01,0.06670719981193543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,32,power_law_1.01,0.6383679866790771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,32,power_law_1.01,0.06552960276603699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,32,power_law_1.01,0.8545408248901367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,32,power_law_1.01,0.06741120219230652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,32,power_law_1.01,1.3088640213012694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,32,power_law_1.01,0.06776319742202759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,32,power_law_1.01,0.06712960004806519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,32,power_law_1.01,2.5091968536376954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,32,power_law_1.01,0.06904320120811462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,32,power_law_1.01,0.07322239875793457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,32,power_law_1.01,0.06754559874534607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,32,power_law_1.01,0.07548159956932068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,32,power_law_1.01,0.06951040029525757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,32,power_law_1.01,0.07960320115089417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,32,power_law_1.01,0.05338240265846252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,32,power_law_1.01,0.09379199743270875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,32,power_law_1.01,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,32,power_law_1.01,0.10389120578765869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,32,power_law_1.01,0.06474239826202392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,32,power_law_1.01,0.12679680585861205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,32,power_law_1.01,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,32,power_law_1.01,0.15264639854431153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,32,power_law_1.01,0.066048002243042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,32,power_law_1.01,0.1928447961807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,32,power_law_1.01,0.06910079717636108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,32,power_law_1.01,0.22770559787750244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,32,power_law_1.01,0.06756479740142822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,32,power_law_1.01,0.32124159336090086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,32,power_law_1.01,0.06855679750442505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,32,power_law_1.01,0.3930047988891602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,32,power_law_1.01,0.06978560090065003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,32,power_law_1.01,0.5702079772949219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,32,power_law_1.01,0.06819199919700622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,32,power_law_1.01,0.7471871852874756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,32,power_law_1.01,0.06995840072631836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,32,power_law_1.01,0.07068799734115601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,32,power_law_1.01,0.946713638305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,32,power_law_1.01,0.0750976026058197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,32,power_law_1.01,1.4622400283813477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,32,power_law_1.01,0.07601919770240784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,32,power_law_1.01,2.8333696365356444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,32,power_law_1.01,0.0780672013759613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,32,power_law_1.01,0.08410879969596863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.01,0.10243200063705445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.01,0.11648000478744507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.01,0.1401087999343872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.01,0.17380479574203492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.01,0.19905920028686525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.01,0.27414400577545167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.01,0.33990399837493895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.01,0.4850304126739502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.01,0.6355072021484375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,32,power_law_1.01,0.8030655860900879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,32,power_law_1.01,1.1689599990844726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,32,power_law_1.01,2.4073856353759764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,32,power_law_1.01,0.02173440009355545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,32,power_law_1.01,0.022815999388694764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,32,power_law_1.01,0.022124800086021423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,32,power_law_1.01,0.021561600267887115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,32,power_law_1.01,0.02383359968662262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,32,power_law_1.01,0.02988159954547882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,32,power_law_1.01,0.05002880096435547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,32,power_law_1.01,0.05016319751739502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,32,power_law_1.01,0.050246399641036985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,32,power_law_1.2,0.07775999903678894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,32,power_law_1.01,0.04594559967517853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,32,power_law_1.01,0.046982398629188536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,32,power_law_1.2,0.0654911994934082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,32,power_law_1.01,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,32,power_law_1.2,0.06352639794349671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,32,power_law_1.01,0.04709759950637817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,32,power_law_1.2,0.0698303997516632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,32,power_law_1.01,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,32,power_law_1.2,0.07611520290374756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,32,power_law_1.01,0.0470335990190506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,32,power_law_1.2,0.07477759718894958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,32,power_law_1.01,0.04705280065536499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,32,power_law_1.2,0.0824512004852295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,32,power_law_1.01,0.04611839950084686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,32,power_law_1.2,0.08047360181808472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,32,power_law_1.01,0.049798399209976196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,32,power_law_1.2,0.0815936028957367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.01,0.052236801385879515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,32,power_law_1.2,0.07983999848365783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.01,0.06558719873428345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,32,power_law_1.2,0.08244479894638061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.01,0.08329600095748901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,32,power_law_1.2,0.0833728015422821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.01,0.1004480004310608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,32,power_law_1.2,0.08408960103988647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.01,0.1230463981628418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,32,power_law_1.2,0.0841152012348175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.01,0.1590016007423401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,32,power_law_1.2,0.08804479837417603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,32,power_law_1.2,0.09303680062294006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.01,0.19730559587478638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,32,power_law_1.2,0.09477760195732117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.01,0.3004096031188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,32,power_law_1.2,0.10408320426940917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.01,0.37188479900360105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,32,power_law_1.2,0.11098239421844483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,32,power_law_1.01,0.46508159637451174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,32,power_law_1.2,0.06142079830169678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,32,power_law_1.2,0.13512959480285644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,32,power_law_1.01,0.7663551807403565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,32,power_law_1.2,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,32,power_law_1.2,0.1625216007232666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,32,power_law_1.2,0.05320320129394531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,32,power_law_1.01,1.412281608581543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,32,power_law_1.2,0.20850560665130616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,32,power_law_1.2,0.0562175989151001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,32,power_law_1.2,0.24101760387420654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,32,power_law_1.2,0.05998079776763916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,32,power_law_1.2,0.3544255971908569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,32,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,32,power_law_1.2,0.4280255794525146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,32,power_law_1.2,0.06406400203704835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,32,power_law_1.2,0.6167808055877686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,32,power_law_1.2,0.06462079882621766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,32,power_law_1.2,0.920076847076416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,32,power_law_1.2,0.06647040247917176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,32,power_law_1.2,1.0060223579406737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,32,power_law_1.2,0.06546559929847717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,32,power_law_1.2,0.06961920261383056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,32,power_law_1.2,0.06648319959640503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,32,power_law_1.2,1.8547775268554687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,32,power_law_1.2,0.05750399827957153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,32,power_law_1.2,0.06748160123825073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,32,power_law_1.2,0.05555840134620667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,32,power_law_1.2,4.272800064086914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,32,power_law_1.2,0.06124160289764404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,32,power_law_1.2,0.0639680027961731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,32,power_law_1.2,0.0651968002319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,32,power_law_1.2,0.06803200244903565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,32,power_law_1.2,0.07050880193710327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,32,power_law_1.2,0.06822400093078614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,32,power_law_1.2,0.07071359753608704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,32,power_law_1.2,0.06937599778175355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,32,power_law_1.2,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,32,power_law_1.2,0.0709824025630951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,32,power_law_1.2,0.06999679803848266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,32,power_law_1.2,0.07554559707641602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,32,power_law_1.2,0.07816960215568543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,32,power_law_1.2,0.08095359802246094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,32,power_law_1.2,0.09372159838676453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.2,0.10612479448318482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.2,0.12012799978256225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.2,0.14383360147476196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.2,0.18743040561676025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.2,0.217574405670166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.2,0.33299200534820556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.2,0.3960383892059326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.2,0.5940351963043213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.2,0.7669504165649415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,32,power_law_1.2,0.9691904067993165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,32,power_law_1.2,1.8307968139648438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,32,power_law_1.2,3.5896831512451173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,32,power_law_1.2,0.021772800385951994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,32,power_law_1.2,0.022060799598693847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,32,power_law_1.2,0.02232320010662079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,32,power_law_1.2,0.02221439927816391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,32,power_law_1.2,0.023647999763488768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,32,power_law_1.2,0.030092799663543703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,32,power_law_1.2,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,32,power_law_1.2,0.05061759948730469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,32,power_law_1.2,0.049958398938179015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,32,power_law_1.2,0.04612480103969574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,32,power_law_1.2,0.04746879935264588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,16,balanced,0.054197331269582115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,16,balanced,0.055349335074424744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,32,power_law_1.2,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,16,balanced,0.057589332262674965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,32,power_law_1.2,0.04757120013237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,16,balanced,0.06325866778691609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,32,power_law_1.2,0.04780800044536591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,16,balanced,0.07326933244864146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,16,balanced,0.08294933537642162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,32,power_law_1.2,0.044556799530982974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,16,balanced,0.08321600159009297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,32,power_law_1.2,0.047014400362968445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,16,balanced,0.08384533723195393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,32,power_law_1.2,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,16,balanced,0.08074666559696198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,32,power_law_1.2,0.04986239969730377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,16,balanced,0.08138133088747661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.2,0.05449600219726562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,16,balanced,0.08281066517035167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,16,balanced,0.0817386656999588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.2,0.06936320066452026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,16,balanced,0.08283733328183492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.2,0.08313599824905396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,16,balanced,0.08536000053087871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.2,0.10970879793167114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,16,balanced,0.08776000142097473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.2,0.1303231954574585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,16,balanced,0.09019733468691508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.2,0.1766592025756836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,16,balanced,0.09387200077374776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.2,0.21398398876190186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,32,power_law_1.2,0.06661760210990905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.2,0.3250175952911377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,16,balanced,0.0976106623808543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,32,power_law_1.2,0.06876800060272217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.2,0.4083199977874756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,16,balanced,0.10204799969991048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,32,power_law_1.2,0.0732159972190857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,32,power_law_1.2,0.5717504024505615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,32,power_law_1.2,0.07670400142669678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,16,balanced,0.11313600341478984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,32,power_law_1.2,0.8532928466796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,32,power_law_1.2,0.08022400140762329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,32,power_law_1.2,0.0964031994342804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,32,power_law_1.2,1.6799360275268556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,16,balanced,0.12363732854525249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,32,power_law_1.2,0.11107200384140015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,32,power_law_1.2,0.13576960563659668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,16,balanced,0.048197334011395775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,32,power_law_1.2,0.1555008053779602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,16,balanced,0.0489333321650823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,16,balanced,0.14963733156522116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,16,balanced,0.05091733237107595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,32,power_law_1.2,0.20930559635162355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,16,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,32,power_law_1.2,0.2532480001449585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,16,balanced,0.06302933394908905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,16,balanced,0.06568533182144165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,32,power_law_1.2,0.3552511930465698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,16,balanced,0.06702933212121327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,16,balanced,0.1658560037612915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,32,power_law_1.2,0.44304637908935546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,16,balanced,0.06783999999364217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,16,balanced,0.06941866874694824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,32,power_law_1.2,0.6631552219390869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,16,balanced,0.06514133512973785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,32,power_law_1.2,0.8687552452087403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,16,balanced,0.07057600220044453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,16,balanced,0.0689333329598109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,32,power_law_1.2,1.162720012664795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,16,balanced,0.06673066814740498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,16,balanced,0.21062399943669638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,32,power_law_1.2,2.2254528045654296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,16,balanced,0.07182933390140533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,16,balanced,0.07294400036334991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,32,power_law_1.2,5.144595336914063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,16,balanced,0.07733866572380066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,16,balanced,0.07954666515191396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,16,balanced,0.08924266695976257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,16,balanced,0.263482669989268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,16,balanced,0.09573866923650105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,16,balanced,0.11990933616956075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,16,balanced,0.13897599776585898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,16,balanced,0.17773866653442383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,16,balanced,0.3439626693725586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,16,balanced,0.20709333817164102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,16,balanced,0.27769599358240765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,16,balanced,0.4413439830144246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,16,balanced,0.35022934277852374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,16,balanced,0.481765349706014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,16,balanced,0.5143200159072876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,16,balanced,0.622437318166097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,16,balanced,0.7542773087819418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,16,balanced,0.7924373149871826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,16,balanced,1.170954704284668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,16,power_law_1.01,0.07489920258522034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,16,power_law_1.01,0.07619199752807618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,16,power_law_1.01,0.05978239774703979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,16,power_law_1.01,0.07230079770088196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,16,power_law_1.01,0.07124480009078979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,16,power_law_1.01,0.07674239873886109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,16,power_law_1.01,0.07874559760093688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,16,power_law_1.01,0.08110719919204712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,16,power_law_1.01,0.08005759716033936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,16,power_law_1.01,0.08154240250587463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,16,power_law_1.01,0.08298239707946778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,16,power_law_1.01,0.08170880079269409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,16,power_law_1.01,0.08595200181007386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,16,power_law_1.01,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,16,power_law_1.01,0.08432639837265014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,16,power_law_1.01,0.06418560147285461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,16,power_law_1.01,0.08967679738998413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,16,power_law_1.01,0.05262079834938049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,16,power_law_1.01,0.09162240028381348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,16,power_law_1.01,0.058406400680541995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,16,power_law_1.01,0.09539200067520141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,16,power_law_1.01,0.061273598670959474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,16,power_law_1.01,0.10372480154037475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,16,power_law_1.01,0.06533120274543762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,16,power_law_1.01,0.06702719926834107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,16,power_law_1.01,0.11326080560684204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,16,power_law_1.01,0.06763520240783691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,16,power_law_1.01,0.12792960405349732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,16,power_law_1.01,0.06714239716529846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,16,power_law_1.01,0.14504319429397583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,16,power_law_1.01,0.06832640171051026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,16,power_law_1.01,0.18378880023956298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,16,power_law_1.01,0.0685375988483429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,16,power_law_1.01,0.2171072006225586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,16,power_law_1.01,0.06974719762802124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,16,power_law_1.01,0.2791167974472046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,16,power_law_1.01,0.06981120109558106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,16,power_law_1.01,0.3838912010192871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,16,balanced,1.5320693651835124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,16,power_law_1.01,0.07199360132217407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,16,power_law_1.01,0.5119935989379882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,16,power_law_1.01,0.07586560249328614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,16,power_law_1.01,0.5996352195739746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,16,power_law_1.01,0.07948799729347229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,16,power_law_1.01,0.7604351997375488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,16,power_law_1.01,0.08483840227127075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,16,power_law_1.01,0.09889919757843017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,16,power_law_1.01,1.2113471984863282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,16,power_law_1.01,0.11217279434204101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,16,power_law_1.01,2.4219648361206056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,16,balanced,2.29422394434611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,16,power_law_1.01,0.13464319705963135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,16,power_law_1.01,0.1600383996963501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,16,power_law_1.01,0.20440959930419922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,16,power_law_1.01,0.24394240379333496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,16,power_law_1.01,0.3411520004272461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,16,power_law_1.01,0.41716480255126953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,16,power_law_1.01,0.6745344161987304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,16,power_law_1.01,0.8038911819458008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,16,power_law_1.01,1.0138879776000977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,16,power_law_1.01,1.6953535079956055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,16,power_law_1.01,3.2719425201416015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,16,power_law_1.2,0.07728000283241272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,16,power_law_1.2,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,16,power_law_1.2,0.06670079827308655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,16,power_law_1.2,0.06969599723815918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,16,power_law_1.2,0.07192959785461425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,16,power_law_1.2,0.07951359748840332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,16,power_law_1.2,0.0639680027961731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,16,power_law_1.2,0.08050559759140015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,16,power_law_1.2,0.052108800411224364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,16,power_law_1.2,0.08193280100822449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,8,balanced,0.05242133140563965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,16,power_law_1.2,0.05288959741592407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,16,power_law_1.2,0.08172799944877625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,8,balanced,0.056943997740745544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,16,power_law_1.2,0.05619199872016907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,16,power_law_1.2,0.08327680230140685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,8,balanced,0.059232001503308616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,16,power_law_1.2,0.059564799070358276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,16,power_law_1.2,0.08236799836158752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,8,balanced,0.06555733581384023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,16,power_law_1.2,0.06446080207824707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,16,power_law_1.2,0.08394880294799804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,8,balanced,0.0788800021012624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,8,balanced,0.0886346697807312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,16,power_law_1.2,0.06635519862174988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,16,power_law_1.2,0.085452800989151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,8,balanced,0.09046399593353271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,16,power_law_1.2,0.0675711989402771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,16,power_law_1.2,0.08654720187187195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,8,balanced,0.08957866827646892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,16,power_law_1.2,0.06652799844741822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,16,power_law_1.2,0.09069439768791199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,8,balanced,0.04990399877230326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,8,balanced,0.0888320008913676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,16,power_law_1.2,0.06581760048866273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,8,balanced,0.0528106689453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,16,power_law_1.2,0.09393919706344604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,8,balanced,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,8,balanced,0.05362133185068766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,16,power_law_1.2,0.06930559873580933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,16,power_law_1.2,0.09700480103492737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,8,balanced,0.08939199646313985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,8,balanced,0.06322133541107178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,16,power_law_1.2,0.0683456003665924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,16,power_law_1.2,0.10607999563217163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,8,balanced,0.0904373327891032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,8,balanced,0.06983999907970428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,16,power_law_1.2,0.06705920100212097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,16,power_law_1.2,0.11355520486831665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,8,balanced,0.08935466408729553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,8,balanced,0.0761706680059433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,16,power_law_1.2,0.0723904013633728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,8,balanced,0.07376533250013988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,8,balanced,0.0899679958820343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,16,power_law_1.2,0.13509119749069215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,8,balanced,0.07681066791216533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,16,power_law_1.2,0.07717120051383972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,16,power_law_1.2,0.14833920001983641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,8,balanced,0.09408533573150635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,8,balanced,0.075013334552447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,16,power_law_1.2,0.0838271975517273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,16,power_law_1.2,0.1922816038131714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,8,balanced,0.09595732887585957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,8,balanced,0.07693333427111308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,16,power_law_1.2,0.09004799723625183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,16,power_law_1.2,0.2483135938644409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,8,balanced,0.09980266292889912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,8,balanced,0.07482133309046428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,16,power_law_1.2,0.10117119550704956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,16,power_law_1.2,0.315391993522644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,8,balanced,0.07469866673151652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,8,balanced,0.10663466652234395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,16,power_law_1.2,0.11649279594421387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,8,balanced,0.07880533238252004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,16,power_law_1.2,0.4071040153503418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,16,power_law_1.2,0.13697279691696168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,8,balanced,0.1125440001487732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,16,power_law_1.2,0.5889984130859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,16,power_law_1.2,0.16339839696884156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,16,power_law_1.2,0.7523776054382324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,8,balanced,0.12451733152071635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,16,power_law_1.2,0.22047359943389894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,16,power_law_1.2,0.8816191673278808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,16,power_law_1.2,0.2864703893661499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,16,power_law_1.2,1.3495936393737793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,16,power_law_1.2,0.38896639347076417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,8,balanced,0.13875733812650046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,16,power_law_1.2,0.5103231906890869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,16,power_law_1.2,3.6361087799072265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,16,power_law_1.2,0.7554048061370849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,8,balanced,0.1740000049273173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,16,power_law_1.2,0.9020352363586426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,16,power_law_1.2,1.2404352188110352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,16,power_law_1.2,1.8290496826171876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,8,balanced,0.193557341893514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,16,power_law_1.2,4.100537490844727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,8,balanced,0.267797331015269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,8,balanced,0.3243040045102437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,8,balanced,0.43214933077494305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,8,balanced,0.5620160102844238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,8,balanced,0.6774506568908691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,8,balanced,1.0487626393636067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,8,power_law_1.01,0.06801279783248901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,8,power_law_1.01,0.07220479846000671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,8,power_law_1.01,0.06478080153465271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,8,power_law_1.01,0.07222399711608887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,8,power_law_1.01,0.0741375982761383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,8,power_law_1.01,0.08429440259933471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,8,power_law_1.01,0.08574720025062561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,8,power_law_1.01,0.061478400230407716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,8,power_law_1.01,0.08712319731712341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,8,power_law_1.01,0.06211199760437012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,8,power_law_1.01,0.08744959831237793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,8,power_law_1.01,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,8,power_law_1.01,0.08687360286712646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,8,power_law_1.01,0.0606656014919281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,8,power_law_1.01,0.08577280044555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,8,power_law_1.01,0.06331520080566407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,8,power_law_1.01,0.08750720024108886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,8,power_law_1.01,0.07071359753608704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,8,power_law_1.01,0.0908352017402649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,8,power_law_1.01,0.07002239823341369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,8,power_law_1.01,0.0920639991760254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,8,power_law_1.01,0.0726207971572876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,8,power_law_1.01,0.09793919920921326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,8,power_law_1.01,0.07214080095291138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,8,power_law_1.01,0.0975488007068634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,8,power_law_1.01,0.07297919988632202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,8,balanced,0.07713066538174947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,8,power_law_1.01,0.10271999835968018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,8,power_law_1.01,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,8,balanced,0.08391466736793518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,8,power_law_1.01,0.11393920183181763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,8,power_law_1.01,0.07504640221595764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,8,balanced,0.0849173367023468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,8,power_law_1.01,0.12467199563980103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,8,power_law_1.01,0.07557119727134705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,8,balanced,0.0885759989420573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,8,power_law_1.01,0.0800000011920929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,8,balanced,0.09876267115275066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,8,power_law_1.01,0.08721920251846313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,8,power_law_1.01,0.09112319946289063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,8,balanced,0.11328533291816711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,8,power_law_1.01,0.09573119878768921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,8,power_law_1.01,0.10918400287628174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,8,balanced,2.0420640309651694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,8,balanced,0.13506133357683817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,8,power_law_1.01,0.12704639434814452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,8,power_law_1.01,0.15411839485168458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,8,balanced,0.15108266472816467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,8,power_law_1.01,0.17307519912719727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,8,power_law_1.01,0.23208320140838623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,8,balanced,0.20205867290496826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,8,power_law_1.01,0.28523519039154055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,8,power_law_1.01,0.3876863956451416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,8,power_law_1.01,0.5008831977844238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,8,balanced,0.23570134242375693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,8,power_law_1.01,0.6943999767303467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,8,power_law_1.01,0.9380224227905274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,8,power_law_1.01,1.1838208198547364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,8,power_law_1.01,1.851705551147461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,8,balanced,0.328821341196696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,8,power_law_1.01,3.303059387207031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,8,balanced,0.41150399049123126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,8,balanced,0.5685653289159139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,8,balanced,0.7444746494293213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,8,balanced,0.9060800075531006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,8,balanced,1.4072160720825195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,8,power_law_1.2,0.06908159852027893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,8,power_law_1.2,0.05979520082473755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,8,power_law_1.2,0.060038399696350095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,8,power_law_1.2,0.06922879815101624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,8,power_law_1.2,0.07423999905586243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,8,power_law_1.2,0.08293759822845459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,8,power_law_1.2,0.08654720187187195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,8,power_law_1.2,0.08983039855957031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,8,power_law_1.2,0.08650879859924317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,8,power_law_1.2,0.060710400342941284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,8,power_law_1.2,0.088755202293396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,8,power_law_1.2,0.05440000295639038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,8,power_law_1.2,0.08947200179100037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,8,power_law_1.01,0.14001280069351196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,8,power_law_1.2,0.0543936014175415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,8,power_law_1.2,0.09065600037574768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,8,power_law_1.01,0.15671039819717408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,8,power_law_1.2,0.05996800065040588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,8,power_law_1.2,0.09079679846763611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,8,power_law_1.01,0.2035775899887085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,8,power_law_1.2,0.060211199522018435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,8,power_law_1.2,0.0934719979763031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,8,power_law_1.01,0.24483199119567872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,8,power_law_1.2,0.07128319740295411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,8,power_law_1.2,0.09901440143585205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,8,power_law_1.01,0.33403520584106444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,8,balanced,2.7867199579874673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,8,power_law_1.2,0.0703935980796814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,8,power_law_1.2,0.10218240022659301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,8,power_law_1.01,0.40844159126281737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,8,power_law_1.2,0.07212799787521362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,8,power_law_1.2,0.10833280086517334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,8,power_law_1.01,0.5874879837036133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,8,power_law_1.2,0.077183997631073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,8,power_law_1.2,0.11347839832305909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,8,power_law_1.01,0.7108799934387207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,8,power_law_1.2,0.07425280213356018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,8,power_law_1.2,0.1296064019203186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,8,power_law_1.01,1.0026047706604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,8,power_law_1.2,0.07512320280075073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,8,power_law_1.2,0.14931199550628663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,8,power_law_1.01,1.5019840240478515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,8,power_law_1.2,0.07713279724121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,8,power_law_1.2,0.16248960494995118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,8,power_law_1.2,0.0753216028213501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,8,power_law_1.2,0.22152318954467773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,8,power_law_1.01,2.8356096267700197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,8,power_law_1.2,0.08056319952011108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,8,power_law_1.2,0.2660928010940552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,8,power_law_1.2,0.08822399973869324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,8,power_law_1.2,0.3520832061767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,8,power_law_1.2,0.0948032021522522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,8,power_law_1.2,0.4758080005645752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,8,power_law_1.2,0.09563519954681396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,8,power_law_1.2,0.7339968204498291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,8,power_law_1.2,0.11475839614868164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,8,power_law_1.2,0.866214370727539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,8,power_law_1.2,0.12767360210418702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,8,power_law_1.2,1.0904895782470703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,8,power_law_1.2,0.15300480127334595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,8,power_law_1.2,1.6842943191528321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,8,power_law_1.2,0.17923840284347534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,8,power_law_1.2,0.24437758922576905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,8,power_law_1.2,3.7304576873779296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,8,power_law_1.2,0.29689600467681887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,8,power_law_1.2,0.43546881675720217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,8,power_law_1.2,0.5613696098327636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,8,power_law_1.2,0.7966080188751221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,8,power_law_1.2,1.0324480056762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,8,power_law_1.2,1.4561344146728517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,8,power_law_1.2,2.053331184387207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,8,power_law_1.2,4.527462387084961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,4,balanced,0.05463466544946035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,4,balanced,0.0591893345117569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,4,balanced,0.061893333991368614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,4,balanced,0.07358933488527934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,4,balanced,0.09397866328557332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,4,balanced,0.05226666728655497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,4,balanced,0.10528533657391866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,4,balanced,0.053488001227378845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,4,balanced,0.10449600219726562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,4,balanced,0.05725333094596863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,4,balanced,0.10726400216420491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,4,balanced,0.06542400022347768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,4,balanced,0.07990399996439616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,4,balanced,0.10406933228174846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,4,balanced,0.0886293351650238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,4,balanced,0.10403733452161153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,4,balanced,0.08992000420888265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,4,balanced,0.10145599643389384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,4,balanced,0.08763200044631958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,4,balanced,0.10550399621327718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,4,balanced,0.08852799733479817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,4,balanced,0.10570666193962097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,4,balanced,0.08860266208648682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,4,balanced,0.10658133029937744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,4,balanced,0.08733333150545756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,4,balanced,0.10983999570210774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,4,balanced,0.0902400016784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,4,balanced,0.1133013367652893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,4,balanced,0.0918880005677541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,4,balanced,0.1160426636536916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,4,balanced,0.0935093363126119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,4,balanced,0.10069866975148518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,4,balanced,0.1267466644446055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,4,balanced,0.10282133022944133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,4,balanced,0.1350986659526825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,4,balanced,0.10801600416501363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,4,power_law_1.01,0.06598399877548218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,4,balanced,0.12633599837621054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,4,balanced,0.15280532836914062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,4,power_law_1.01,0.06535680294036865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,4,balanced,0.13667733470598856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,4,power_law_1.01,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,4,balanced,0.1677173376083374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,4,power_law_1.01,0.07331839799880982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,4,balanced,0.16328000028928122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,4,power_law_1.01,0.08265600204467774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,4,power_law_1.01,0.09547520279884339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,4,balanced,0.23886932929356894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,4,balanced,0.18603734175364176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,4,power_law_1.01,0.09833599925041199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,4,power_law_1.01,0.06058880090713501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,4,power_law_1.01,0.09901440143585205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,4,power_law_1.01,0.06272640228271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,4,power_law_1.01,0.10151040554046631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,4,balanced,0.26129599412282306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,4,balanced,0.27141332626342773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,4,power_law_1.01,0.0574400007724762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,4,power_law_1.01,0.10123519897460938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,4,power_law_1.01,0.06429439783096313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,4,power_law_1.01,0.10300159454345703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,4,power_law_1.01,0.0701312005519867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,4,power_law_1.01,0.10363520383834839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,4,balanced,0.30482665697733563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,4,power_law_1.01,0.07811840176582337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,4,power_law_1.01,0.10412160158157349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,4,power_law_1.01,0.10586880445480347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,4,power_law_1.01,0.08124160170555114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,4,balanced,0.36101333300272626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,4,power_law_1.01,0.1118783950805664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,4,power_law_1.01,0.08418560028076172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,4,balanced,0.4223039944966634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,4,power_law_1.01,0.1141759991645813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,4,power_law_1.01,0.08520960211753845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,4,power_law_1.01,0.11882239580154419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,4,power_law_1.01,0.08570240139961242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,4,power_law_1.01,0.13240959644317626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,4,power_law_1.01,0.08857600092887878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,4,power_law_1.01,0.14600319862365724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,4,balanced,0.46348798274993896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,4,power_law_1.01,0.09081599712371827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,4,power_law_1.01,0.16842880249023437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,4,power_law_1.01,0.0901311993598938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,4,power_law_1.01,0.1912384033203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,4,power_law_1.01,0.09423360228538513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,4,balanced,0.5509440104166666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,4,power_law_1.01,0.270963191986084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,4,power_law_1.01,0.10362880229949951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,4,power_law_1.01,0.3168447971343994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,4,power_law_1.01,0.10743680000305175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,4,power_law_1.01,0.42414078712463377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,4,power_law_1.01,0.11519360542297363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,4,power_law_1.01,0.1320639967918396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,4,power_law_1.01,0.5164608001708985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,4,power_law_1.01,0.15077120065689087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,4,power_law_1.01,0.7500351905822754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,4,balanced,0.6161919832229614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,4,power_law_1.01,0.17726720571517945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,4,power_law_1.01,0.9826623916625976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,4,balanced,0.7590346336364746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,4,power_law_1.01,0.22278399467468263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,4,power_law_1.01,1.1748160362243651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,4,power_law_1.01,0.2878848075866699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,4,power_law_1.01,1.986969566345215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,4,power_law_1.01,0.3480448007583618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,4,power_law_1.01,0.48946561813354494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,4,power_law_1.01,3.6366912841796877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,4,power_law_1.01,0.613920021057129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,4,power_law_1.01,0.9061120033264161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,4,power_law_1.01,1.1702336311340331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,4,balanced,0.8075573444366455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,4,power_law_1.01,1.3971136093139649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,4,balanced,0.9992960294087728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,4,power_law_1.01,2.33111686706543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,4,power_law_1.01,4.74376335144043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,4,balanced,0.9677759806315104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,4,balanced,1.2174346446990967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,4,balanced,1.9005813598632812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,4,balanced,1.5063306490580242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,4,balanced,3.7773707707722983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,4,balanced,3.015461285909017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,4,power_law_1.2,0.06503679752349853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,4,power_law_1.2,0.058303999900817874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,4,power_law_1.2,0.05363839864730835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,4,power_law_1.2,0.06104320287704468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,4,power_law_1.2,0.05607039928436279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,4,power_law_1.2,0.06577280163764954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,4,power_law_1.2,0.06247040033340454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,4,power_law_1.2,0.07470719814300537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,4,power_law_1.2,0.06863999962806702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,4,power_law_1.2,0.07886080145835876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,4,power_law_1.2,0.07946239709854126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,4,power_law_1.2,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,4,power_law_1.2,0.08536319732666016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,4,power_law_1.2,0.10013439655303955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,4,power_law_1.2,0.0838271975517273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,4,power_law_1.2,0.10025600194931031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,4,power_law_1.2,0.08620799779891967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,4,power_law_1.2,0.10323200225830079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,4,power_law_1.2,0.0834879994392395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,4,power_law_1.2,0.10162559747695923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,4,power_law_1.2,0.08959360122680664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,4,power_law_1.2,0.10216319561004639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,4,power_law_1.2,0.0929472029209137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,4,power_law_1.2,0.10412800312042236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,4,power_law_1.2,0.09268479943275451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,4,power_law_1.2,0.10250879526138305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,4,power_law_1.2,0.09448320269584656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,4,power_law_1.2,0.1070207953453064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,4,power_law_1.2,0.10457600355148315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,4,power_law_1.2,0.11357439756393432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,4,power_law_1.2,0.10616320371627808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,4,power_law_1.2,0.11512960195541382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,4,power_law_1.2,0.11610879898071289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,4,power_law_1.2,0.12400640249252319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,4,power_law_1.2,0.13375359773635864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,4,power_law_1.2,0.1363711953163147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,4,power_law_1.2,0.15267839431762695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,4,power_law_1.2,0.1491520047187805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,4,power_law_1.2,0.18443520069122316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,4,power_law_1.2,0.17468160390853882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,4,power_law_1.2,0.20291199684143066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,4,power_law_1.2,0.21772160530090331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,4,power_law_1.2,0.28952960968017577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,4,power_law_1.2,0.27594881057739257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,4,power_law_1.2,0.3635456085205078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,4,power_law_1.2,0.30196480751037597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,4,power_law_1.2,0.5084671974182129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,4,power_law_1.2,0.4176703929901123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,4,power_law_1.2,0.6451263904571534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,4,power_law_1.2,0.5709248065948487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,4,power_law_1.2,1.016089630126953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,4,power_law_1.2,0.8739456176757813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,4,power_law_1.2,1.2459391593933105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,4,power_law_1.2,1.1237695693969727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,4,power_law_1.2,1.5257856369018554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,4,power_law_1.2,1.5297792434692383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,4,power_law_1.2,2.4620479583740233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,4,power_law_1.2,2.203455924987793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,4,power_law_1.2,5.213081741333008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,4,power_law_1.2,4.485964965820313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,128,balanced,0.08909866213798523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,128,balanced,0.08601599931716919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,128,balanced,0.09011733531951904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,128,balanced,0.09379200140635173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,128,balanced,0.09223999579747517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,128,balanced,0.10282666484514873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,128,balanced,0.12164266904195149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,128,balanced,0.11565333604812622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,128,balanced,0.11440533399581909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,128,balanced,0.06443200012048085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,128,balanced,0.06654933094978333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,128,balanced,0.11782399813334148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,128,balanced,0.06516266862551372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,128,balanced,0.1288266678651174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,128,balanced,0.06586666901906331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,128,balanced,0.12036266922950745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,128,balanced,0.067930668592453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,128,balanced,0.12519466876983643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,128,balanced,0.07748800019423167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,128,balanced,0.08826667070388794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,128,balanced,0.13590400417645773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,128,balanced,0.086709330479304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,128,balanced,0.13142399986584982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,128,balanced,0.0890826682249705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,128,balanced,0.14235732952753702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,128,balanced,0.08839999636014302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,128,balanced,0.09428266684214275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,128,balanced,0.1437120040257772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,128,balanced,0.09342933694521587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,128,balanced,0.1402079959710439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,128,balanced,0.0935093363126119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,128,balanced,0.09737599889437358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,128,balanced,0.15165332953135172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,128,balanced,0.10187199711799622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,128,balanced,0.10173333684603374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,128,balanced,0.1612266699473063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,128,balanced,0.11082667112350464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,128,balanced,0.12005333105723064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,128,balanced,0.16500799854596457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,128,balanced,0.1267466644446055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,128,balanced,0.15019200245539346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,128,balanced,0.18828266859054565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,128,balanced,0.16994667053222656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,128,balanced,0.2119413415590922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,128,balanced,0.20743467410405478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,128,balanced,0.2479520042737325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,128,balanced,0.26480533679326373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,128,balanced,0.32941333452860516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,128,balanced,0.3492319981257121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,128,balanced,0.4262506564458211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,128,balanced,0.06454933186372121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,128,balanced,0.5825119813283285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,128,balanced,0.061664000153541565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,128,balanced,0.06202666461467743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,128,balanced,0.44332265853881836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,128,power_law_1.01,0.1115455985069275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,128,balanced,0.06346666812896729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,128,balanced,0.06757333377997081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,128,power_law_1.01,0.11061760187149047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,128,balanced,0.07308800021807353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,128,balanced,0.08576533198356628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,128,power_law_1.01,0.1025920033454895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,128,balanced,0.09021332859992981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,128,power_law_1.01,0.09673600196838379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,128,balanced,0.08565333485603333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,128,balanced,0.08896000186602275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,128,power_law_1.01,0.11036159992218017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,128,balanced,0.08980266253153484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,128,balanced,0.09143466750780742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,128,power_law_1.01,0.10803200006484985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,128,balanced,0.7445120016733805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,128,balanced,0.0881866713364919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,128,power_law_1.01,0.11006079912185669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,128,balanced,0.0916426678498586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,128,power_law_1.01,0.11054719686508178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,128,balanced,0.10417600472768147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,128,power_law_1.01,0.12216960191726685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,128,balanced,0.09570133686065674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,128,balanced,0.1079200009504954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,128,power_law_1.01,0.11469440460205078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,128,balanced,0.5527306795120239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,128,balanced,0.10878400007883708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,128,power_law_1.01,0.11663999557495117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,128,power_law_1.01,0.1229632019996643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,128,balanced,0.12216533223787944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,128,power_law_1.01,0.12260479927062988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,128,balanced,0.1340053379535675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,128,power_law_1.01,0.12760319709777831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,128,power_law_1.01,0.12721279859542847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,128,balanced,0.149125337600708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,128,balanced,0.9298773606618246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,128,power_law_1.01,0.1276927947998047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,128,power_law_1.01,0.13068159818649291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,128,balanced,0.17668267091115317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,128,power_law_1.01,0.1331264019012451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,128,power_law_1.01,0.1388159990310669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,128,power_law_1.01,0.1729920029640198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,128,balanced,0.2050079902013143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,128,power_law_1.01,0.19127039909362792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,128,balanced,0.7023893197377523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,128,power_law_1.01,0.24145278930664063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,128,power_law_1.01,0.2973311901092529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,128,balanced,0.2653546730677287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,128,power_law_1.01,0.4047935962677002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,128,power_law_1.01,0.4820352077484131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,128,power_law_1.01,0.6728256225585938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,128,power_law_1.01,0.8479935646057128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,128,balanced,0.34349334239959717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,128,power_law_1.01,1.1052543640136718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,128,power_law_1.01,1.747270393371582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,128,balanced,1.4302825927734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,128,power_law_1.01,3.5631168365478514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,128,balanced,0.4531466563542684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,128,balanced,1.0508586565653484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,128,balanced,0.586464007695516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,128,balanced,0.7265973091125488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,128,balanced,2.771669387817383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,128,balanced,1.1121546427408855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,128,balanced,1.9728693962097168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,128,balanced,2.128906726837158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,128,power_law_1.01,0.07759360074996949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,128,power_law_1.01,0.07544959783554077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,128,power_law_1.01,0.07402240037918091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,128,power_law_1.01,0.07438079714775085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,128,power_law_1.01,0.07924479842185975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,128,power_law_1.01,0.08376320004463196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,128,power_law_1.01,0.0802944004535675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,128,power_law_1.01,0.08887040019035339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,128,power_law_1.01,0.08951680064201355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,128,power_law_1.01,0.0876800000667572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,128,power_law_1.01,0.08892800211906433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,128,power_law_1.01,0.08768640160560608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,128,power_law_1.01,0.09310719966888428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,128,power_law_1.01,0.09420160055160523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,128,power_law_1.01,0.09933440089225769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,128,power_law_1.01,0.10337920188903808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,128,power_law_1.01,0.10675200223922729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,128,power_law_1.01,0.11280640363693237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,128,power_law_1.01,0.12725759744644166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,128,power_law_1.01,0.15768959522247314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,128,power_law_1.01,0.18585599660873414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,128,power_law_1.01,0.23150079250335692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,128,power_law_1.01,0.2893951892852783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,128,power_law_1.01,0.4114560127258301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,128,power_law_1.01,0.5116032123565674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,128,power_law_1.01,0.7435520172119141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,128,power_law_1.01,0.9504128456115722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,128,power_law_1.01,1.1241727828979493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,128,power_law_1.01,1.7525888442993165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,128,power_law_1.01,0.08099200129508972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,128,power_law_1.01,0.08036479949951172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,128,power_law_1.01,3.5485374450683596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,128,power_law_1.01,0.08655359745025634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,128,power_law_1.01,0.07423999905586243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,128,power_law_1.01,0.07756800055503846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,128,power_law_1.01,0.07793279886245727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,128,power_law_1.01,0.08640639781951905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,128,power_law_1.01,0.08634880185127258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,128,power_law_1.01,0.0831488013267517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,128,power_law_1.01,0.08329600095748901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,128,power_law_1.01,0.08309760093688964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,128,power_law_1.01,0.08476160168647766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,128,power_law_1.2,0.10474879741668701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,128,power_law_1.01,0.08781440258026123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,128,power_law_1.2,0.11132800579071045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,128,power_law_1.2,0.10687359571456909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,128,power_law_1.01,0.08479359745979309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,128,power_law_1.2,0.0989247977733612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,128,power_law_1.01,0.09543039798736572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,128,power_law_1.2,0.11427839994430541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,128,power_law_1.01,0.09617919921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,128,power_law_1.2,0.11098239421844483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,128,power_law_1.01,0.09966080188751221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,128,power_law_1.2,0.10794880390167236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,128,power_law_1.01,0.10730880498886108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,128,power_law_1.2,0.11667200326919555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,128,power_law_1.01,0.12067199945449829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,128,power_law_1.2,0.11767040491104126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,128,power_law_1.01,0.14381439685821534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,128,power_law_1.2,0.12717440128326415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,128,power_law_1.01,0.16215039491653443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,128,power_law_1.2,0.10854400396347046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,128,power_law_1.01,0.20770559310913086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,128,power_law_1.2,0.1214400053024292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,128,power_law_1.01,0.25022718906402586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,128,power_law_1.2,0.12300159931182861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,128,power_law_1.01,0.33832321166992185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,128,power_law_1.2,0.12090239524841309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,128,power_law_1.01,0.4132351875305176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,128,power_law_1.2,0.1267840027809143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,128,power_law_1.01,0.5716351985931396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,128,power_law_1.2,0.12039680480957031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,128,power_law_1.01,0.7585343837738037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,128,power_law_1.2,0.12463359832763672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,128,power_law_1.01,0.916710376739502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,128,power_law_1.2,0.13764480352401734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,128,power_law_1.01,1.4445055961608886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,128,power_law_1.2,0.15256320238113402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,128,power_law_1.2,0.19587839841842652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,128,power_law_1.01,2.7881023406982424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,128,power_law_1.2,0.2190335988998413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,128,power_law_1.2,0.2678591966629028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,128,power_law_1.2,0.07944959998130799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,128,power_law_1.2,0.08395519852638245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,128,power_law_1.2,0.3361151933670044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,128,power_law_1.2,0.07845119833946228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,128,power_law_1.2,0.4452159881591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,128,power_law_1.2,0.07361279726028443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,128,power_law_1.2,0.5628672122955323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,128,power_law_1.2,0.07473919987678528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,128,power_law_1.2,0.9234880447387696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,128,power_law_1.2,0.08107519745826722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,128,power_law_1.2,1.163327980041504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,128,power_law_1.2,0.08460800051689148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,128,power_law_1.2,1.4828607559204101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,128,power_law_1.2,0.08694400191307068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,128,power_law_1.2,2.2071744918823244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,128,power_law_1.2,0.08519039750099182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,128,power_law_1.2,0.08858240246772767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,128,power_law_1.2,4.548979187011719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,128,power_law_1.2,0.08903679847717286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,128,power_law_1.2,0.09570559859275818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,128,power_law_1.2,0.09345279932022095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,128,power_law_1.2,0.09068160057067871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,128,power_law_1.2,0.10183039903640748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,128,power_law_1.2,0.10158079862594604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,128,power_law_1.2,0.10651520490646363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,128,power_law_1.2,0.11923199892044067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,128,power_law_1.2,0.13397760391235353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,128,power_law_1.2,0.16828160285949706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,128,power_law_1.2,0.1958847999572754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,128,power_law_1.2,0.2522495985031128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,128,power_law_1.2,0.3066368103027344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,128,power_law_1.2,0.4268671989440918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,128,power_law_1.2,0.5663424015045166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,128,power_law_1.2,0.8611647605895996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,128,power_law_1.2,1.0592448234558105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,128,power_law_1.2,1.3277055740356445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,128,power_law_1.2,2.132236862182617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,128,power_law_1.2,4.2646526336669925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,128,power_law_1.2,0.08027520179748535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,128,power_law_1.2,0.0859391987323761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,128,power_law_1.2,0.07850880026817322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,128,power_law_1.2,0.07604479789733887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,128,power_law_1.2,0.07638400197029113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,128,power_law_1.2,0.07615360021591186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,128,power_law_1.2,0.08223360180854797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,128,power_law_1.2,0.07956479787826538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,128,power_law_1.2,0.08265600204467774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,128,power_law_1.2,0.08472959995269776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,128,power_law_1.2,0.08498560190200806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,128,power_law_1.2,0.08609920144081115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,128,power_law_1.2,0.08866559863090515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,128,power_law_1.2,0.08647680282592773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,128,power_law_1.2,0.0946175992488861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,128,power_law_1.2,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,128,power_law_1.2,0.09710720181465149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,128,power_law_1.2,0.11240320205688477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,128,power_law_1.2,0.12262400388717651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,128,power_law_1.2,0.14855040311813356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,128,power_law_1.2,0.1793023943901062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,128,power_law_1.2,0.22458879947662352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,128,power_law_1.2,0.27992959022521974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,128,power_law_1.2,0.36862080097198485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,128,power_law_1.2,0.4934400081634521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,128,power_law_1.2,0.6436351776123047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,128,power_law_1.2,0.8928192138671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,128,power_law_1.2,1.1350336074829102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,128,power_law_1.2,1.7951040267944336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,128,power_law_1.2,3.546771240234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,64,balanced,0.06676266590754192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,64,balanced,0.07010666529337566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,64,balanced,0.06901333232720692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,64,balanced,0.07158400118350983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,64,balanced,0.08091733356316884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,64,balanced,0.09986666838328044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,64,balanced,0.10917333761850993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,64,balanced,0.11254933476448059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,64,balanced,0.11954133709271748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,64,balanced,0.1186240017414093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,64,balanced,0.11461866895357768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,64,balanced,0.11665067076683044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,64,balanced,0.05444799860318502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,64,balanced,0.11564266681671143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,64,balanced,0.05749866863091787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,64,balanced,0.11351999640464783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,64,balanced,0.05715199808279673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,64,balanced,0.12618133425712585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,64,balanced,0.06078400214513143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,64,balanced,0.0631573349237442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,64,balanced,0.12191466490427653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,64,balanced,0.07401599983374278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,64,balanced,0.1334826648235321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,64,balanced,0.08403199911117554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,64,balanced,0.08417066931724548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,64,balanced,0.14008000493049622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,64,balanced,0.09168000022570293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,64,balanced,0.08826667070388794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,64,balanced,0.1328159968058268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,64,balanced,0.0888266662756602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,64,balanced,0.09017067154248555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,64,balanced,0.15040533741315207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,64,balanced,0.09092266360918681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,64,balanced,0.09227733810742696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,64,balanced,0.10406399766604106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,64,balanced,0.16345066825548807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,64,balanced,0.09974933664004008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,64,balanced,0.10804266730944316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,64,balanced,0.18742400407791138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,64,balanced,0.11867733796437581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,64,balanced,0.12562666336695352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,64,balanced,0.21253865957260132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,64,balanced,0.14697600404421488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,64,balanced,0.16553599635759988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,64,balanced,0.26453866561253864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,64,balanced,0.2104640007019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,64,balanced,0.2492213249206543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,64,balanced,0.3519573211669922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,64,balanced,0.056218668818473816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,64,balanced,0.05749866863091787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,64,balanced,0.05506666501363119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,64,balanced,0.05805333455403646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,64,balanced,0.0671253353357315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,64,balanced,0.3313973347345988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,64,balanced,0.07764266431331635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,64,balanced,0.08955732981363933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,64,balanced,0.0836853285630544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,64,balanced,0.08590933680534363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,64,balanced,0.08762666583061218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,64,balanced,0.44679466883341473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,64,balanced,0.08918933073679607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,64,balanced,0.08565866947174072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,64,balanced,0.08762666583061218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,64,balanced,0.42902934551239014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,64,balanced,0.08964266379674275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,64,balanced,0.09823999802271526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,64,balanced,0.09563733140627544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,64,balanced,0.10547733306884766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,64,balanced,0.10936533411343892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,64,balanced,0.11867200334866841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,64,balanced,0.1346826652685801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,64,balanced,0.5579573313395182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,64,balanced,0.5893653233846029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,64,balanced,0.14939733346303305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,64,balanced,0.17742933829625449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,64,balanced,0.20815465847651163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,64,balanced,0.26929599046707153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,64,balanced,0.7557493050893148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,64,balanced,0.7041066487630209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,64,balanced,0.3424746592839559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,64,balanced,0.46143468221028644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,64,balanced,0.937935988108317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,64,power_law_1.01,0.1122048020362854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,64,power_law_1.01,0.10771199464797973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,64,power_law_1.01,0.10430719852447509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,64,power_law_1.01,0.08260480165481568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,64,balanced,1.0571626822153728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,64,power_law_1.01,0.0923520028591156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,64,balanced,0.5948799848556519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,64,power_law_1.01,0.09934080243110657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,64,power_law_1.01,0.10024960041046142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,64,power_law_1.01,0.10826239585876465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,64,power_law_1.01,0.11085439920425415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,64,power_law_1.01,0.10954240560531617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,64,power_law_1.01,0.11136000156402588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,64,power_law_1.01,0.10657279491424561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,64,power_law_1.01,0.1093824028968811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,64,power_law_1.01,0.11472640037536622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,64,balanced,0.737173318862915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,64,power_law_1.01,0.12021759748458863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,64,power_law_1.01,0.1149183988571167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,64,power_law_1.01,0.11737600564956666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,64,balanced,1.4409279823303223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,64,power_law_1.01,0.12778240442276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,64,power_law_1.01,0.13160320520401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,64,power_law_1.01,0.15270400047302246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,64,power_law_1.01,0.18028160333633422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,64,power_law_1.01,0.2240959882736206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,64,power_law_1.01,0.2557440042495728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,64,power_law_1.01,0.36881279945373535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,64,power_law_1.01,0.4355775833129883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,64,power_law_1.01,0.6098495960235596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,64,power_law_1.01,0.792793607711792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,64,power_law_1.01,0.947270393371582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,64,balanced,1.1294666926066081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,64,power_law_1.01,1.520639991760254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,64,power_law_1.01,2.7960704803466796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,64,balanced,2.0095200538635254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,64,balanced,2.8224960962931314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,64,balanced,2.1845547358194985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,64,power_law_1.01,0.08244479894638061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,64,power_law_1.01,0.07923840284347534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,64,power_law_1.01,0.07767040133476258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,64,power_law_1.01,0.06558719873428345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,64,power_law_1.01,0.06741120219230652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,64,power_law_1.01,0.08369920253753663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,64,power_law_1.01,0.07727360129356384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,64,power_law_1.01,0.08837119936943054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,64,power_law_1.01,0.08081920146942138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,64,power_law_1.01,0.08299520015716552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,64,power_law_1.01,0.07885439991950989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,64,power_law_1.01,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,64,power_law_1.01,0.08162559866905213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,64,power_law_1.01,0.07386879920959473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,64,power_law_1.01,0.07104640007019043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,64,power_law_1.01,0.0801472008228302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,64,power_law_1.01,0.07758079767227173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,64,power_law_1.01,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,64,power_law_1.01,0.0828607976436615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,64,power_law_1.01,0.08724480271339416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,64,power_law_1.01,0.07533439993858337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,64,power_law_1.01,0.08930559754371643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,64,power_law_1.01,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,64,power_law_1.01,0.08937600255012512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,64,power_law_1.01,0.08208640217781067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,64,power_law_1.01,0.0946175992488861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,64,power_law_1.01,0.08558719754219055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,64,power_law_1.01,0.09489279985427856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,64,power_law_1.01,0.08849920034408569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,64,power_law_1.01,0.10735360383987427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,64,power_law_1.01,0.08727040290832519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,64,power_law_1.01,0.11166080236434936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,64,power_law_1.01,0.09411839842796325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,64,power_law_1.01,0.12389119863510131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,64,power_law_1.2,0.11179519891738891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,64,power_law_1.01,0.09525759816169739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,64,power_law_1.01,0.14864640235900878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,64,power_law_1.2,0.10520960092544555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,64,power_law_1.01,0.09739519953727722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,64,power_law_1.2,0.09969279766082764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,64,power_law_1.01,0.1765247941017151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,64,power_law_1.01,0.10830719470977783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,64,power_law_1.2,0.08769279718399048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,64,power_law_1.01,0.23372159004211426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,64,power_law_1.01,0.11893759965896607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,64,power_law_1.2,0.08885120153427124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,64,power_law_1.01,0.2784640073776245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,64,power_law_1.01,0.13895679712295533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,64,power_law_1.2,0.09847040176391601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,64,power_law_1.01,0.38224639892578127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,64,power_law_1.2,0.08160640001296997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,64,power_law_1.01,0.155840003490448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,64,power_law_1.2,0.09944959878921508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,64,power_law_1.01,0.498905611038208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,64,power_law_1.2,0.07793279886245727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,64,power_law_1.01,0.20383999347686768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,64,power_law_1.2,0.10382720232009887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,64,power_law_1.01,0.6895808219909668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,64,power_law_1.2,0.07386239767074584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,64,power_law_1.01,0.24408960342407227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,64,power_law_1.2,0.10671360492706299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,64,power_law_1.01,0.8844799995422363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,64,power_law_1.01,0.32038400173187254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,64,power_law_1.2,0.10687359571456909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,64,power_law_1.2,0.06258559823036194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,64,power_law_1.01,1.0852864265441895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,64,power_law_1.2,0.10782719850540161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,64,power_law_1.01,0.398963189125061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,64,power_law_1.2,0.07052159905433655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,64,power_law_1.01,1.7540864944458008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,64,power_law_1.2,0.11213439702987671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,64,power_law_1.01,0.5596735954284668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,64,power_law_1.2,0.07404800057411194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,64,power_law_1.2,0.10872960090637207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,64,power_law_1.01,0.7292799949645996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,64,power_law_1.2,0.07791360020637512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,64,power_law_1.01,3.529996871948242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,64,power_law_1.2,0.10694400072097779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,64,power_law_1.01,0.898630428314209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,64,power_law_1.2,0.08296319842338562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,64,power_law_1.2,0.1137279987335205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,64,power_law_1.2,0.08263040184974671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,64,power_law_1.01,1.3766271591186523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,64,power_law_1.2,0.11571840047836304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,64,power_law_1.2,0.08257920145988465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,64,power_law_1.01,2.7717504501342773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,64,power_law_1.2,0.12095999717712402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,64,power_law_1.2,0.08569599986076355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,64,power_law_1.2,0.1321023941040039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,64,power_law_1.2,0.08825600147247314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,64,power_law_1.2,0.13943040370941162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,64,power_law_1.2,0.08762239813804626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,64,power_law_1.2,0.16606080532073975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,64,power_law_1.2,0.08574079871177673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,64,power_law_1.2,0.20222721099853516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,64,power_law_1.2,0.09703680276870727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,64,power_law_1.2,0.2504447937011719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,64,power_law_1.2,0.09950720071792603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,64,power_law_1.2,0.2965183973312378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,64,power_law_1.2,0.10566400289535523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,64,power_law_1.2,0.3819967985153198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,64,power_law_1.2,0.11633919477462769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,64,power_law_1.2,0.5025087833404541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,64,power_law_1.2,0.12755199670791625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,64,power_law_1.2,0.7678912162780762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,64,power_law_1.2,0.16166399717330932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,64,power_law_1.2,0.9621760368347168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,64,power_law_1.2,0.18940160274505616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,64,power_law_1.2,1.1096832275390625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,64,power_law_1.2,0.243833589553833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,64,power_law_1.2,1.7148351669311523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,64,power_law_1.2,0.3066112041473389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,64,power_law_1.2,3.6747905731201174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,64,power_law_1.2,0.40320639610290526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,64,power_law_1.2,0.5224959850311279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,64,power_law_1.2,0.7531199932098389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,64,power_law_1.2,1.0172991752624512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,64,power_law_1.2,1.2302783966064452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,64,power_law_1.2,2.1189247131347657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,64,power_law_1.2,4.182579040527344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,32,balanced,0.05878399809201559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,64,power_law_1.2,0.088128000497818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,32,balanced,0.06098133325576782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,32,balanced,0.0591893345117569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,64,power_law_1.2,0.084934401512146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,32,balanced,0.06592000027497609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,64,power_law_1.2,0.06985599994659424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,32,balanced,0.07606400052706401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,32,balanced,0.09482133388519287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,64,power_law_1.2,0.06661120057106018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,32,balanced,0.10851200421651204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,32,balanced,0.10479467113812764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,64,power_law_1.2,0.07086079716682434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,32,balanced,0.10995200276374817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,64,power_law_1.2,0.07279360294342041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,32,balanced,0.11115200320879619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,32,balanced,0.11101333300272624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,64,power_law_1.2,0.07654399871826172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,32,balanced,0.1127946674823761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,64,power_law_1.2,0.08279039859771728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,32,balanced,0.10991467038790385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,32,balanced,0.1181813379128774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,64,power_law_1.2,0.080595201253891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,32,balanced,0.11251200238863628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,64,power_law_1.2,0.08032000064849854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,32,balanced,0.11411733428637187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,64,power_law_1.2,0.08300160169601441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,32,balanced,0.1204906702041626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,64,power_law_1.2,0.08469759821891784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,32,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,32,balanced,0.12654399871826172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,32,balanced,0.05366399884223938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,64,power_law_1.2,0.08273280262947083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,32,balanced,0.0529013325770696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,32,balanced,0.1336373289426168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,64,power_law_1.2,0.08970879912376403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,32,balanced,0.05827199916044871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,32,balanced,0.05919999877611796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,64,power_law_1.2,0.09323520064353943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,32,balanced,0.07334400216738383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,32,balanced,0.1472640037536621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,32,balanced,0.08779199918111165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,64,power_law_1.2,0.0934656023979187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,32,balanced,0.08576533198356628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,64,power_law_1.2,0.09835519790649414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,32,balanced,0.16142400105794272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,32,balanced,0.08594133456548055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,32,balanced,0.08733333150545756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,64,power_law_1.2,0.11419520378112794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,32,balanced,0.08738666772842407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,64,power_law_1.2,0.12182400226593018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,32,balanced,0.08971200386683147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,32,balanced,0.19115199645360312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,32,balanced,0.09197333455085754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,64,power_law_1.2,0.1422719955444336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,32,balanced,0.09499733646710713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,64,power_law_1.2,0.16762880086898804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,32,balanced,0.10149332880973816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,64,power_law_1.2,0.20488319396972657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,32,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,32,balanced,0.10365866621335347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,32,balanced,0.21638933817545572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,32,balanced,0.053360000252723694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,64,power_law_1.2,0.2598912000656128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,32,balanced,0.10703999797503154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,32,balanced,0.05328000088532766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,32,balanced,0.059157331784566246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,64,power_law_1.2,0.3560703992843628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,32,balanced,0.1172160009543101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,32,balanced,0.06666666766007741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,32,balanced,0.07550399998823802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,64,power_law_1.2,0.45060482025146487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,32,balanced,0.12794133027394614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,32,balanced,0.09402666489283244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,32,balanced,0.2730773289998372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,64,power_law_1.2,0.6486207962036132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,32,balanced,0.09474666913350423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,32,balanced,0.15089600284894308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,32,balanced,0.09347732861836751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,64,power_law_1.2,0.8025471687316894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,32,balanced,0.09141866366068523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,32,balanced,0.0925439993540446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,64,power_law_1.2,1.0075584411621095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,32,balanced,0.16929600636164346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,32,balanced,0.0897759993871053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,64,power_law_1.2,1.5896127700805665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,32,balanced,0.09125866492589314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,32,balanced,0.35555732250213623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,32,balanced,0.09659199913342793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,32,balanced,0.21414399147033691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,64,power_law_1.2,3.1935552597045898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,32,balanced,0.09888533751169841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,32,balanced,0.09780266880989075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,32,balanced,0.10646933317184448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,32,balanced,0.2555413246154785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,32,balanced,0.11277332901954651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,32,balanced,0.12059199810028076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,32,balanced,0.46288001537323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,32,balanced,0.13685866196950278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,32,balanced,0.34136001269022626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,32,balanced,0.15556800365447998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,32,balanced,0.18746666113535562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,32,balanced,0.4430026610692342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,32,balanced,0.21855467557907104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,32,balanced,0.5825333197911581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,32,balanced,0.2836853265762329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,32,balanced,0.608351985613505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,32,balanced,0.3585653305053711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,32,balanced,0.784495989481608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,32,balanced,0.7215733528137207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,32,balanced,0.48714665571848553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,32,balanced,0.627946654955546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,32,balanced,0.9675626754760742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,32,balanced,1.0913279851277669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,32,balanced,0.7764960130055746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,32,balanced,1.4979359308878581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,32,balanced,1.187941312789917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,32,balanced,2.0979092915852866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,32,power_law_1.01,0.09633920192718506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,32,power_law_1.01,0.1109503984451294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,32,power_law_1.01,0.10666879415512084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,32,power_law_1.01,0.07744640111923218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,32,balanced,2.9292160669962564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,32,power_law_1.01,0.08444160223007202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,32,power_law_1.01,0.09548159837722778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,32,power_law_1.01,0.09447680115699768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,32,power_law_1.01,0.10156160593032837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,32,power_law_1.01,0.09879040122032165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,32,power_law_1.01,0.0985152006149292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,32,power_law_1.01,0.1021183967590332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,32,balanced,2.317082722981771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,32,power_law_1.01,0.10339200496673584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,32,power_law_1.01,0.10767999887466431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,32,power_law_1.01,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,32,power_law_1.01,0.1120576024055481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,32,power_law_1.01,0.11722240447998047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,32,power_law_1.01,0.11809279918670654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,32,power_law_1.01,0.12781440019607543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,32,power_law_1.01,0.13415039777755738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,32,power_law_1.01,0.15336320400238038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,32,power_law_1.01,0.1690816044807434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,32,power_law_1.01,0.20624001026153566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,32,power_law_1.01,0.25959680080413816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,32,power_law_1.01,0.3358400106430054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,32,power_law_1.01,0.42342400550842285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,32,power_law_1.01,0.582809591293335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,32,power_law_1.01,0.7939072132110596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,32,power_law_1.01,0.9040767669677734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,32,power_law_1.01,1.4630911827087403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,32,power_law_1.01,2.6380031585693358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,32,power_law_1.01,0.07422080039978027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,32,power_law_1.01,0.08371840119361877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,32,power_law_1.01,0.08193280100822449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,32,power_law_1.01,0.060185599327087405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,32,power_law_1.01,0.08204159736633301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,32,power_law_1.01,0.06854400038719177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,32,power_law_1.01,0.09397760033607483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,32,power_law_1.01,0.08687360286712646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,32,power_law_1.01,0.07433599829673768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,32,power_law_1.01,0.07748479843139648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,32,power_law_1.01,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,32,power_law_1.01,0.07968639731407165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,32,power_law_1.01,0.06699519753456115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,32,power_law_1.01,0.07907840013504028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,32,power_law_1.01,0.07178239822387696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,32,power_law_1.01,0.07674239873886109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,32,power_law_1.01,0.0811456024646759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,32,power_law_1.01,0.08310400247573853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,32,power_law_1.01,0.08155519962310791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,32,power_law_1.01,0.08396160006523132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,32,power_law_1.01,0.08184959888458251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,32,power_law_1.01,0.08469759821891784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,32,power_law_1.01,0.08897280097007751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,32,power_law_1.2,0.0936191976070404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,32,power_law_1.01,0.08496639728546143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,32,power_law_1.01,0.08777599930763244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,32,power_law_1.2,0.10567679405212402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,32,power_law_1.01,0.08902400135993957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,32,power_law_1.01,0.09718400239944458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,32,power_law_1.2,0.09637759923934937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,32,power_law_1.01,0.08499839901924133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,32,power_law_1.01,0.09685760140419006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,32,power_law_1.2,0.07669119834899903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,32,power_law_1.01,0.08957440257072449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,32,power_law_1.01,0.10333440303802491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,32,power_law_1.2,0.08191999793052673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,32,power_law_1.01,0.09682559967041016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,32,power_law_1.01,0.11905280351638795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,32,power_law_1.2,0.08595839738845826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,32,power_law_1.01,0.09993600249290466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,32,power_law_1.01,0.12686079740524292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,32,power_law_1.01,0.10104320049285889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,32,power_law_1.2,0.09016960263252258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,32,power_law_1.01,0.1538815975189209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,32,power_law_1.01,0.11310080289840699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,32,power_law_1.2,0.09887359738349914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,32,power_law_1.01,0.17798399925231934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,32,power_law_1.01,0.12454400062561036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,32,power_law_1.2,0.09782400131225585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,32,power_law_1.01,0.2351167917251587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,32,power_law_1.01,0.14702719449996948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,32,power_law_1.2,0.09652479887008666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,32,power_law_1.01,0.28361599445343016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,32,power_law_1.01,0.16295039653778076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,32,power_law_1.2,0.1005247950553894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,32,power_law_1.01,0.3837440013885498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,32,power_law_1.01,0.20611839294433593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,32,power_law_1.2,0.10409599542617798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,32,power_law_1.01,0.49258880615234374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,32,power_law_1.2,0.10183039903640748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,32,power_law_1.01,0.24715518951416016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,32,power_law_1.01,0.6917952060699463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,32,power_law_1.2,0.10273280143737792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,32,power_law_1.01,0.3305599927902222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,32,power_law_1.01,0.9095616340637207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,32,power_law_1.2,0.11022720336914063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,32,power_law_1.01,0.4088768005371094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,32,power_law_1.01,1.1291199684143067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,32,power_law_1.2,0.11189759969711303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,32,power_law_1.01,0.5726272106170655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,32,power_law_1.01,1.8052288055419923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,32,power_law_1.2,0.11594239473342896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,32,power_law_1.01,0.7327360153198242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,32,power_law_1.2,0.12910720109939575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,32,power_law_1.01,3.403603363037109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,32,power_law_1.01,0.9128447532653808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,32,power_law_1.2,0.1394368052482605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,32,power_law_1.01,1.4098239898681642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,32,power_law_1.2,0.15644160509109498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,32,power_law_1.01,2.736115264892578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,32,power_law_1.2,0.18993279933929444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,32,power_law_1.2,0.2352832078933716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,32,power_law_1.2,0.2770944118499756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,32,power_law_1.2,0.07428479790687562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,32,power_law_1.2,0.37114880084991453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,32,power_law_1.2,0.08665599822998046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,32,power_law_1.2,0.4620672225952148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,32,power_law_1.2,0.08044800162315369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,32,power_law_1.2,0.613651180267334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,32,power_law_1.2,0.0667967975139618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,32,power_law_1.2,0.07009919881820678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,32,power_law_1.2,0.8390912055969239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,32,power_law_1.2,0.07195519804954528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,32,power_law_1.2,0.9682944297790528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,32,power_law_1.2,0.0784704029560089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,32,power_law_1.2,1.5718720436096192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,32,power_law_1.2,0.08256000280380249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,32,power_law_1.2,3.2360958099365233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,32,power_law_1.2,0.08194559812545776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,32,power_law_1.2,0.08190720081329346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,32,power_law_1.2,0.08043519854545593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,32,power_law_1.2,0.0871999979019165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,32,power_law_1.2,0.0877951979637146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,32,power_law_1.2,0.09107199907302857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,32,power_law_1.2,0.09676160216331482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,32,power_law_1.2,0.09993600249290466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,32,power_law_1.2,0.10659199953079224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,32,power_law_1.2,0.11946879625320435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,32,power_law_1.2,0.1318336009979248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,32,power_law_1.2,0.16011519432067872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,32,power_law_1.2,0.18719359636306762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,32,power_law_1.2,0.2447551965713501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,32,power_law_1.2,0.3004800081253052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,32,power_law_1.2,0.42174720764160156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,32,power_law_1.2,0.5063551902770996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,32,power_law_1.2,0.7760447978973388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,32,power_law_1.2,0.9964032173156738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,32,power_law_1.2,1.2547840118408202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,32,power_law_1.2,1.8908992767333985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,32,power_law_1.2,3.639443206787109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,16,balanced,0.0587360014518102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,32,power_law_1.2,0.08053119778633118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,16,balanced,0.05932799975077311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,16,balanced,0.062314664324124656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,32,power_law_1.2,0.0915008008480072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,16,balanced,0.06734933455785115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,32,power_law_1.2,0.08431360125541687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,16,balanced,0.07994133234024048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,16,balanced,0.09981866677602132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,32,power_law_1.2,0.06376960277557372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,16,balanced,0.11387200156847636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,32,power_law_1.2,0.06601600050926208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,16,balanced,0.11303466558456421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,16,balanced,0.11503466963768005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,32,power_law_1.2,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,16,balanced,0.11449600259462993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,32,power_law_1.2,0.0772159993648529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,16,balanced,0.11051199833552043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,32,power_law_1.2,0.07934719920158387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,16,balanced,0.11577600240707397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,16,balanced,0.1104159951210022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,32,power_law_1.2,0.0788096010684967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,16,balanced,0.05060799916585287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,16,balanced,0.11538133025169373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,32,power_law_1.2,0.0844223976135254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,16,balanced,0.0517439991235733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,16,balanced,0.05342933535575867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,16,balanced,0.11868799726168315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,32,power_law_1.2,0.08551040291786194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,16,balanced,0.06099733213583628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,16,balanced,0.12154666582743327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,32,power_law_1.2,0.08506879806518555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,16,balanced,0.06321066617965698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,16,balanced,0.07783466577529907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,16,balanced,0.1283466617266337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,32,power_law_1.2,0.08463360071182251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,16,balanced,0.09435199697812398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,16,balanced,0.13751999537150064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,32,power_law_1.2,0.09029120206832886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,16,balanced,0.09223467111587524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,16,balanced,0.09618666768074036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,32,power_law_1.2,0.09651200175285339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,16,balanced,0.13783466815948486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,16,balanced,0.09751466910044353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,32,power_law_1.2,0.09681280255317688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,16,balanced,0.0937546690305074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,16,balanced,0.15599999825159708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,32,power_law_1.2,0.1007040023803711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,16,balanced,0.09546132882436116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,16,balanced,0.09822400410970052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,32,power_law_1.2,0.11543680429458618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,16,balanced,0.09899733463923137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,16,balanced,0.16985599199930826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,32,power_law_1.2,0.1251456022262573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,16,balanced,0.1095360020796458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,32,power_law_1.2,0.14597760438919066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,16,balanced,0.10729599992434184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,32,power_law_1.2,0.16935039758682252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,16,balanced,0.20173333088556925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,16,balanced,0.11276800433794658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,32,power_law_1.2,0.22984321117401124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,16,balanced,0.12506133317947388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,32,power_law_1.2,0.2627135992050171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,16,balanced,0.13759467005729675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,32,power_law_1.2,0.36846721172332764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,16,balanced,0.2302186687787374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,32,power_law_1.2,0.4503359794616699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,16,balanced,0.16145599881807962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,32,power_law_1.2,0.648364782333374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,16,balanced,0.18296533823013306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,32,power_law_1.2,0.8249792098999024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,32,power_law_1.2,1.0274880409240723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,16,balanced,0.2910240093866984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,32,power_law_1.2,1.6505088806152344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,16,balanced,0.23056000471115112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,32,power_law_1.2,3.3484161376953123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,16,balanced,0.273418664932251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,16,balanced,0.3839893341064453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,16,balanced,0.05464000006516775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,16,balanced,0.36472535133361816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,16,balanced,0.05569066603978475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,16,balanced,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,16,balanced,0.06278933087984721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,16,balanced,0.07178666690985362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,16,balanced,0.08194666604201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,16,balanced,0.10481066505114238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,16,balanced,0.09889599680900574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,16,balanced,0.47701867421468097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,16,balanced,0.09835199515024821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,16,balanced,0.5070826609929403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,16,balanced,0.10228266318639119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,16,balanced,0.10133333007494609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,16,balanced,0.10259733597437541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,16,balanced,0.10019200046857198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,16,balanced,0.10634666681289673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,16,balanced,0.11141332983970642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,16,balanced,0.11126933495203654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,16,balanced,0.659333348274231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,16,balanced,0.11900800466537476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,16,balanced,0.12642133235931396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,16,balanced,0.13642666737238565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,16,balanced,0.15398400028546652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,16,balanced,0.643285314242045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,16,balanced,0.1728960076967875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,16,balanced,0.8538293043772379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,16,balanced,0.2079733411471049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,16,balanced,0.2422986626625061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,16,balanced,0.3110719919204712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,16,balanced,0.7941760222117106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,16,balanced,1.0546613534291585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,16,balanced,0.4010293483734131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,16,balanced,0.5439680020014445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,16,balanced,1.624000072479248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,16,balanced,1.2001813252766926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,16,balanced,0.7006133397420248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,16,balanced,0.8672320048014323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,16,power_law_1.01,0.08094080090522766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,16,power_law_1.01,0.09966080188751221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,16,power_law_1.01,0.110099196434021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,16,power_law_1.01,0.07430400252342224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,16,balanced,1.3346506754557292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,16,power_law_1.01,0.0818560004234314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,16,balanced,3.179706573486328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,16,power_law_1.01,0.09141759872436524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,16,power_law_1.01,0.09304959774017334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,16,power_law_1.01,0.10105600357055664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,16,power_law_1.01,0.10273280143737792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,16,power_law_1.01,0.10077439546585083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,16,balanced,2.3134613037109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,16,power_law_1.01,0.10419199466705323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,16,power_law_1.01,0.10316159725189208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,16,power_law_1.01,0.10346879959106445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,16,power_law_1.01,0.10648319721221924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,16,power_law_1.01,0.1144063949584961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,16,power_law_1.01,0.11704959869384765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,16,power_law_1.01,0.11829760074615478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,16,power_law_1.01,0.1269695997238159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,16,power_law_1.01,0.13867520093917846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,16,power_law_1.01,0.16123520135879515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,16,power_law_1.01,0.17768959999084472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,16,power_law_1.01,0.22888319492340087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,16,power_law_1.01,0.27239038944244387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,16,power_law_1.01,0.3485055923461914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,16,power_law_1.01,0.42740478515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,16,power_law_1.01,0.6166656017303467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,16,power_law_1.01,0.7787327766418457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,16,power_law_1.01,0.9519424438476562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,16,power_law_1.01,1.4865407943725586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,16,power_law_1.01,2.76940803527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,16,balanced,2.601050694783529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,16,power_law_1.01,0.06835839748382569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,16,power_law_1.01,0.0789247989654541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,16,power_law_1.01,0.09038720130920411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,16,power_law_1.01,0.07317119836807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,16,power_law_1.01,0.06334720253944397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,16,power_law_1.01,0.08730880022048951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,16,power_law_1.01,0.07032319903373718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,16,power_law_1.01,0.0996288001537323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,16,power_law_1.01,0.07347840070724487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,16,power_law_1.01,0.06222720146179199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,16,power_law_1.01,0.07728000283241272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,16,power_law_1.01,0.0673088014125824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,16,power_law_1.01,0.0812608003616333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,16,power_law_1.01,0.08634880185127258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,16,power_law_1.01,0.08896639943122864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,16,power_law_1.01,0.08710399866104127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,16,power_law_1.01,0.08954880237579346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,16,power_law_1.01,0.09094399809837342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,16,power_law_1.01,0.09434880018234253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,16,power_law_1.01,0.1028864026069641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,16,power_law_1.01,0.1046336054801941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,16,power_law_1.01,0.11136000156402588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,16,power_law_1.01,0.12236160039901733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,16,power_law_1.01,0.13642879724502563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,16,power_law_1.01,0.1702656030654907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,16,power_law_1.01,0.19215359687805175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,16,power_law_1.01,0.2504512071609497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,16,power_law_1.01,0.31573119163513186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,16,power_law_1.01,0.4213376045227051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,16,power_law_1.01,0.5291007995605469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,16,power_law_1.01,0.7415616035461425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,16,power_law_1.01,0.9373567581176758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,16,power_law_1.01,1.1904128074645997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,16,power_law_1.01,1.8580991744995117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,16,power_law_1.01,3.6269760131835938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,16,power_law_1.2,0.07674239873886109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,16,power_law_1.2,0.06893439888954163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,16,power_law_1.2,0.10243840217590332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,16,power_law_1.2,0.0824832022190094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,16,power_law_1.2,0.0886080026626587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,16,power_law_1.2,0.07541120052337646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,16,power_law_1.2,0.07575680017471313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,16,power_law_1.2,0.0652671992778778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,16,power_law_1.2,0.08612480163574218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,16,power_law_1.2,0.07147520184516906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,16,power_law_1.2,0.08982399702072144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,16,power_law_1.2,0.07327359914779663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,16,power_law_1.2,0.09441919922828675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,16,power_law_1.2,0.09968000054359435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,16,power_law_1.2,0.07633280158042907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,16,power_law_1.2,0.09806079864501953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,16,power_law_1.2,0.0850048005580902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,16,power_law_1.2,0.0987712025642395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,16,power_law_1.2,0.08389120101928711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,16,power_law_1.2,0.1009984016418457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,16,power_law_1.2,0.08560640215873719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,16,power_law_1.2,0.08809599876403809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,16,power_law_1.2,0.10468480587005616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,16,power_law_1.2,0.10744320154190064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,16,power_law_1.2,0.09063040018081665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,16,power_law_1.2,0.10622080564498901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,16,power_law_1.2,0.0923583984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,16,power_law_1.2,0.1142848014831543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,16,power_law_1.2,0.0939199984073639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,16,power_law_1.2,0.116211199760437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,16,power_law_1.2,0.10379519462585449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,16,power_law_1.2,0.10469119548797608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,16,power_law_1.2,0.12157440185546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,16,power_law_1.2,0.1332352042198181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,16,power_law_1.2,0.11397759914398194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,16,power_law_1.2,0.1448192000389099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,16,power_law_1.2,0.12584960460662842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,16,power_law_1.2,0.16784000396728516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,16,power_law_1.2,0.1401087999343872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,16,power_law_1.2,0.18646399974822997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,16,power_law_1.2,0.1701311945915222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,16,power_law_1.2,0.19475200176239013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,16,power_law_1.2,0.24686079025268554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,16,power_law_1.2,0.28166399002075193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,16,power_law_1.2,0.2611840009689331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,16,power_law_1.2,0.38248319625854493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,16,power_law_1.2,0.3188544034957886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,16,power_law_1.2,0.46672639846801756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,16,power_law_1.2,0.4381120204925537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,16,power_law_1.2,0.5629824161529541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,16,power_law_1.2,0.6492095947265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,16,power_law_1.2,0.8279423713684082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,16,power_law_1.2,0.8755200386047364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,16,power_law_1.2,0.9963711738586426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,16,power_law_1.2,0.9950592041015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,16,power_law_1.2,1.2109439849853516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,16,power_law_1.2,1.7215999603271483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,16,power_law_1.2,2.0137088775634764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,16,power_law_1.2,3.5578495025634767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,16,power_law_1.2,4.085420989990235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,16,power_law_1.01,0.07398399710655212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,16,power_law_1.01,0.07990400195121765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,16,power_law_1.01,0.08576639890670776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,16,power_law_1.01,0.08520960211753845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,16,power_law_1.01,0.09109119772911071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,16,power_law_1.01,0.0942911982536316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,16,power_law_1.01,0.09646720290184022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,16,power_law_1.01,0.09439359903335572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,16,power_law_1.01,0.09807999730110169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,16,power_law_1.01,0.10494719743728638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,16,power_law_1.2,0.0712768018245697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,16,power_law_1.01,0.1059648036956787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,16,power_law_1.2,0.08761600255966187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,16,power_law_1.01,0.11144319772720337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,16,power_law_1.2,0.07642239928245545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,16,power_law_1.01,0.12624000310897826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,16,power_law_1.2,0.06344959735870362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,16,power_law_1.01,0.13667839765548706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,16,power_law_1.2,0.06729599833488464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,16,power_law_1.01,0.1595903992652893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,16,power_law_1.2,0.07382400035858154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,16,power_law_1.01,0.17639039754867553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,16,power_law_1.2,0.07746559977531434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,16,power_law_1.01,0.2243135929107666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,16,power_law_1.2,0.08270080089569092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,16,power_law_1.01,0.2761280059814453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,16,power_law_1.2,0.08819839954376221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,16,power_law_1.01,0.36022400856018066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,16,power_law_1.2,0.0856383979320526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,16,power_law_1.2,0.09173759818077087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,16,power_law_1.01,0.45976958274841306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,16,power_law_1.2,0.09075199961662292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,16,power_law_1.01,0.6420735836029052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,16,power_law_1.2,0.0932096004486084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,16,power_law_1.01,0.8392512321472168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,16,power_law_1.2,0.09813759922981262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,16,power_law_1.01,1.0362367630004883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,16,power_law_1.2,0.10622080564498901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,16,power_law_1.01,1.5867456436157226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,16,power_law_1.2,0.10533759593963624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,16,power_law_1.2,0.11216000318527222
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,16,power_law_1.01,3.102547264099121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,16,power_law_1.2,0.1292736053466797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,16,power_law_1.2,0.1388479948043823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,16,power_law_1.2,0.16275839805603026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,16,power_law_1.2,0.1902783989906311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,16,power_law_1.2,0.23237760066986085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,16,power_law_1.2,0.28752000331878663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,16,power_law_1.2,0.3824703931808472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,16,power_law_1.2,0.4870272159576416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,16,power_law_1.2,0.6894336223602295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,16,power_law_1.2,0.9039999961853027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,16,power_law_1.2,1.1554495811462402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,16,power_law_1.2,1.723961639404297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,16,power_law_1.2,3.6402496337890624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,8,balanced,0.05669866502285004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,8,balanced,0.058693334460258484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,8,balanced,0.06308799982070923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,8,balanced,0.07025599976380666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,8,balanced,0.08578667044639587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,8,balanced,0.05044800043106079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,8,balanced,0.1072213351726532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,8,balanced,0.05320000151793162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,8,balanced,0.1257866621017456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,8,balanced,0.05764799813429514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,8,balanced,0.12436800201733907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,8,balanced,0.06316266457239787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,8,balanced,0.06942933301130931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,8,balanced,0.12429866194725037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,8,balanced,0.08897067109743755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,8,balanced,0.1218986709912618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,8,balanced,0.10504000385602315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,8,balanced,0.12267200152079265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,8,balanced,0.10500799616177876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,8,balanced,0.1258240044116974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,8,balanced,0.10318932930628459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,8,balanced,0.12457600235939026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,8,balanced,0.10494400064150493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,8,balanced,0.1253600021203359
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,8,balanced,0.10588266452153523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,8,balanced,0.13038933277130127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,8,balanced,0.10987200339635213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,8,balanced,0.13170133034388223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,8,balanced,0.11036800344785054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,8,balanced,0.1374613344669342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,8,balanced,0.11053867141405742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,8,balanced,0.14760532975196838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,8,balanced,0.11931199828783672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,8,balanced,0.12098133563995361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,8,balanced,0.15705600380897522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,8,balanced,0.128602663675944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,8,balanced,0.17725332578023276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,8,balanced,0.14250666896502176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,8,balanced,0.1949173410733541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,8,balanced,0.1553546686967214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,8,balanced,0.18302400906880698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,8,balanced,0.2316533327102661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,8,balanced,0.2044693430264791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,8,balanced,0.2638453245162964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,8,balanced,0.2576906681060791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,8,balanced,0.3432533343633016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,8,balanced,0.3054933349291484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,8,balanced,0.4570293426513672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,8,balanced,0.41070401668548584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,8,balanced,0.6108800172805786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,8,balanced,0.5470613241195679
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,8,balanced,0.7775946458180746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,8,balanced,0.7525226275126139
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,8,balanced,0.054245332876841225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,8,balanced,0.059024001161257424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,8,balanced,0.06108266611893972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,8,balanced,0.07083733379840851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,8,balanced,0.08051200211048126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,8,balanced,0.10418666402498881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,8,balanced,0.13080533345540366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,8,balanced,0.13523733615875244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,8,balanced,0.13379733761151633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,8,balanced,0.13074666261672974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,8,balanced,0.13343999783198038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,8,balanced,0.954416036605835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,8,balanced,0.13658133149147034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,8,balanced,0.9829066594441732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,8,balanced,0.1300106644630432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,8,balanced,0.14014933506647745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,8,balanced,0.1420906682809194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,8,balanced,0.14410666624704996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,8,balanced,0.15085867047309875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,8,balanced,0.1591146687666575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,8,balanced,0.17018133401870728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,8,power_law_1.01,0.0722495973110199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,8,power_law_1.01,0.08312960267066956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,8,balanced,0.19037866592407227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,8,power_law_1.01,0.10877439975738526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,8,power_law_1.01,0.0762112021446228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,8,balanced,0.2090239922205607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,8,power_law_1.01,0.08318719863891602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,8,power_law_1.01,0.0969983994960785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,8,balanced,0.25305600961049396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,8,power_law_1.01,0.10278400182723998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,8,balanced,1.2101333141326904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,8,power_law_1.01,0.10780800580978393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,8,power_law_1.01,0.10839040279388427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,8,balanced,1.4530612627665203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,8,balanced,0.2906720042228699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,8,power_law_1.01,0.11064319610595703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,8,power_law_1.01,0.11015679836273193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,8,power_law_1.01,0.11632000207901001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,8,power_law_1.01,0.11388800144195557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,8,balanced,0.3797066609064738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,8,power_law_1.01,0.11944960355758667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,8,power_law_1.01,0.12383359670639038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,8,power_law_1.01,0.12447359561920165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,8,power_law_1.01,0.13345919847488402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,8,power_law_1.01,0.14510079622268676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,8,balanced,0.5018879969914755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,8,power_law_1.01,0.15438719987869262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,8,power_law_1.01,0.1802240014076233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,8,power_law_1.01,0.2025279998779297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,8,power_law_1.01,0.2519295930862427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,8,power_law_1.01,0.31157760620117186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,8,power_law_1.01,0.40908160209655764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,8,power_law_1.01,0.5053952217102051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,8,balanced,0.6788319746653239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,8,balanced,1.8670986493428547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,8,power_law_1.01,0.7186367988586426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,8,power_law_1.01,0.8708095550537109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,8,power_law_1.01,1.0845952033996582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,8,power_law_1.01,1.7040000915527345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,8,power_law_1.01,3.3619457244873048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,8,balanced,0.888597329457601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,8,balanced,2.7836586634318032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,8,balanced,1.0912373065948486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,8,balanced,1.6765012741088867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,8,balanced,3.649301211039225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,8,balanced,3.2566986083984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,8,power_law_1.01,0.0650816023349762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,8,power_law_1.01,0.07307519912719726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,8,power_law_1.01,0.08472959995269776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,8,power_law_1.01,0.06726400256156921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,8,power_law_1.01,0.06997759938240052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,8,power_law_1.01,0.08055040240287781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,8,power_law_1.01,0.08469759821891784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,8,power_law_1.01,0.0912447988986969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,8,power_law_1.01,0.09280639886856079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,8,power_law_1.01,0.09416319727897644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,8,power_law_1.01,0.09783679842948914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,8,power_law_1.01,0.09813119769096375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,8,power_law_1.01,0.10247039794921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,8,power_law_1.01,0.10696320533752442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,8,power_law_1.01,0.07236480116844177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,8,power_law_1.01,0.11516799926757812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,8,power_law_1.01,0.0866815984249115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,8,power_law_1.01,0.11826560497283936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,8,power_law_1.01,0.10777599811553955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,8,power_law_1.01,0.12651519775390624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,8,power_law_1.01,0.06709120273590088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,8,power_law_1.01,0.1376512050628662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,8,power_law_1.01,0.08074880242347718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,8,power_law_1.01,0.15523840188980104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,8,power_law_1.01,0.08974080085754395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,8,power_law_1.01,0.19215999841690062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,8,power_law_1.01,0.09897599816322326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,8,power_law_1.01,0.216646409034729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,8,power_law_1.01,0.11014399528503419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,8,power_law_1.01,0.2724607944488525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,8,power_law_1.01,0.11642240285873413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,8,power_law_1.01,0.34288640022277833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,8,power_law_1.01,0.11536639928817749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,8,power_law_1.01,0.46696319580078127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,8,power_law_1.01,0.11872639656066894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,8,power_law_1.01,0.5981247901916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,8,power_law_1.01,0.12123520374298095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,8,power_law_1.01,0.8235584259033203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,8,power_law_1.01,0.12136319875717164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,8,power_law_1.01,1.0756223678588868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,8,power_law_1.01,0.12323199510574341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,8,power_law_1.01,1.3336511611938477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,8,power_law_1.01,0.13616000413894652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,8,power_law_1.01,2.133203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,8,power_law_1.01,0.13559039831161498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,8,power_law_1.01,0.1410431981086731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,8,power_law_1.01,4.215750503540039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,8,power_law_1.01,0.15754879713058473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,8,power_law_1.01,0.17068159580230713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,8,power_law_1.2,0.07324159741401673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,8,power_law_1.01,0.19680000543594361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,8,power_law_1.2,0.08506240248680115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,8,power_law_1.01,0.2244352102279663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,8,power_law_1.2,0.08511999845504761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,8,power_law_1.01,0.27802879810333253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,8,power_law_1.2,0.07685120105743408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,8,power_law_1.01,0.32685439586639403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,8,power_law_1.2,0.08362879753112792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,8,power_law_1.01,0.4369344234466553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,8,power_law_1.2,0.09594240188598632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,8,power_law_1.01,0.5456064224243165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,8,power_law_1.2,0.095769602060318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,8,power_law_1.01,0.7982592105865478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,8,power_law_1.2,0.10444159507751465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,8,power_law_1.2,0.10851199626922607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,8,power_law_1.01,1.028428840637207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,8,power_law_1.2,0.11144319772720337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,8,power_law_1.01,1.268064022064209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,8,power_law_1.2,0.11137280464172364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,8,power_law_1.01,1.9206911087036134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,8,power_law_1.2,0.11169279813766479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,8,power_law_1.01,3.753900909423828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,8,power_law_1.2,0.11476479768753052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,8,power_law_1.2,0.11599359512329102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,8,power_law_1.2,0.12496639490127563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,8,power_law_1.2,0.12666879892349242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,8,power_law_1.2,0.13415679931640626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,8,power_law_1.2,0.15182080268859863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,8,power_law_1.2,0.15864319801330568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,8,power_law_1.2,0.18360960483551025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,8,power_law_1.2,0.2081984043121338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,8,power_law_1.2,0.2662911891937256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,8,power_law_1.2,0.3346175909042358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,8,power_law_1.2,0.4382336139678955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,8,power_law_1.2,0.514521598815918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,8,power_law_1.2,0.7373568058013916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,8,power_law_1.2,0.9890432357788086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,8,power_law_1.2,0.06585599780082703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,8,power_law_1.2,0.07082239985466003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,8,power_law_1.2,1.2853952407836915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,8,power_law_1.2,0.07041280269622803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,8,power_law_1.2,1.8451456069946288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,8,power_law_1.2,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,8,power_law_1.2,3.8122879028320313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,8,power_law_1.2,0.07125120162963867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,8,power_law_1.2,0.07982720136642456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,8,power_law_1.2,0.08409600257873535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,8,power_law_1.2,0.09131519794464112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,8,power_law_1.2,0.09490560293197632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,8,power_law_1.2,0.09785599708557129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,8,power_law_1.2,0.09626880288124084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,8,power_law_1.2,0.10004479885101318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,8,power_law_1.2,0.10128639936447144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,8,power_law_1.2,0.10645120143890381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,8,power_law_1.2,0.11645439863204957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,8,power_law_1.2,0.11860480308532714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,8,power_law_1.2,0.12599040269851686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,8,power_law_1.2,0.14094079732894899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,8,power_law_1.2,0.15924479961395263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,8,power_law_1.2,0.19001599550247192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,8,power_law_1.2,0.22449920177459717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,8,power_law_1.2,0.28418560028076173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,8,power_law_1.2,0.34771840572357177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,8,power_law_1.2,0.49423999786376954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,8,power_law_1.2,0.6090176105499268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,8,power_law_1.2,0.9186495780944824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,8,power_law_1.2,1.146457576751709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,8,power_law_1.2,1.4210240364074707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,8,power_law_1.2,2.361248016357422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,8,power_law_1.2,4.246233749389648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,8,power_law_1.2,0.07218559980392455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,8,power_law_1.2,0.08520320057868958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,8,power_law_1.2,0.07170559763908387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,8,power_law_1.2,0.06893439888954163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,8,power_law_1.2,0.07749760150909424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,8,power_law_1.2,0.0862272024154663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,8,power_law_1.2,0.09705600142478943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,8,power_law_1.2,0.10311679840087891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,8,power_law_1.2,0.11271040439605713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,8,power_law_1.2,0.11483520269393921
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,8,power_law_1.2,0.12004480361938477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,8,power_law_1.2,0.12046719789505005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,8,power_law_1.2,0.11898880004882813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,8,power_law_1.2,0.11932159662246704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,8,power_law_1.2,0.1297279953956604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,8,power_law_1.2,0.13475199937820434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,4,balanced,0.060693333546320595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,8,power_law_1.2,0.1409343957901001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,4,balanced,0.06525866687297821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,4,balanced,0.06778133412202199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,8,power_law_1.2,0.15879039764404296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,4,balanced,0.07753066718578339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,4,balanced,0.09456533193588257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,8,power_law_1.2,0.1725376009941101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,4,balanced,0.130431999762853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,8,power_law_1.2,0.20261759757995607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,4,balanced,0.14342400431632996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,4,balanced,0.14416533708572388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,8,power_law_1.2,0.22171521186828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,4,balanced,0.14545599619547525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,4,balanced,0.1416373352209727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,8,power_law_1.2,0.2791872024536133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,4,balanced,0.14453867077827454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,8,power_law_1.2,0.3382528066635132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,4,balanced,0.1467466652393341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,4,balanced,0.14726932843526205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,8,power_law_1.2,0.460422420501709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,4,balanced,0.1474293371041616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,8,power_law_1.2,0.619923210144043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,4,balanced,0.1577173372109731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,8,power_law_1.2,0.8625280380249023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,4,balanced,0.15730667114257812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,8,power_law_1.2,1.1224512100219726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,4,balanced,0.16689600547154745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,8,power_law_1.2,1.3174015998840332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,4,balanced,0.18039466937383017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,4,balanced,0.054245332876841225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,8,power_law_1.2,2.112031936645508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,4,balanced,0.19362666209538779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,4,balanced,0.05769599974155426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,4,balanced,0.06188266475995382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,8,power_law_1.2,4.128774261474609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,4,balanced,0.07096000015735626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,4,balanced,0.21553067366282144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,4,balanced,0.08392000198364258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,4,balanced,0.11249599854151408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,4,balanced,0.13521066308021545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,4,balanced,0.24041599035263062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,4,balanced,0.1365493337313334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,4,balanced,0.13456533352533975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,4,balanced,0.13730667034784952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,4,balanced,0.13724266489346823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,4,balanced,0.291978657245636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,4,balanced,0.13868266344070435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,4,balanced,0.14122666915257773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,4,balanced,0.1423520048459371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,4,balanced,0.15285866459210715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,4,balanced,0.3417760133743286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,4,balanced,0.15310399731000265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,4,balanced,0.16280532876650491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,4,balanced,0.18012799819310507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,4,balanced,0.19383466243743896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,4,balanced,0.4466079870859782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,4,balanced,0.22940800587336221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,4,balanced,0.2540106574694316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,4,balanced,0.604965329170227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,4,balanced,0.32103466987609863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,4,balanced,0.38397332032521564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,4,balanced,0.813589334487915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,4,balanced,0.5159840186436971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,4,power_law_1.01,0.07208319902420043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,4,power_law_1.01,0.07654399871826172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,4,power_law_1.01,0.09553279876708984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,4,power_law_1.01,0.08104959726333619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,4,power_law_1.01,0.09114879965782166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,4,balanced,0.7011146545410156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,4,power_law_1.01,0.10485759973526002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,4,power_law_1.01,0.11182719469070435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,4,balanced,1.0479093392690022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,4,power_law_1.01,0.12108160257339477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,4,power_law_1.01,0.12284799814224243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,4,power_law_1.01,0.12660479545593262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,4,power_law_1.01,0.12739839553833007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,4,balanced,0.9583040078481039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,4,power_law_1.01,0.13204480409622193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,4,power_law_1.01,0.13572479486465455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,4,power_law_1.01,0.13713279962539673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,4,power_law_1.01,0.14568320512771607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,4,power_law_1.01,0.1503872036933899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,4,power_law_1.01,0.1562880039215088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,4,power_law_1.01,0.17093119621276856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,4,balanced,1.2937653064727783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,4,power_law_1.01,0.18647680282592774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,4,power_law_1.01,0.2209088087081909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,4,balanced,1.2692373593648274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,4,power_law_1.01,0.2501312017440796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,4,power_law_1.01,0.3094912052154541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,4,power_law_1.01,0.3688575983047485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,4,power_law_1.01,0.5027967929840088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,4,power_law_1.01,0.6383103847503662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,4,power_law_1.01,0.8669887542724609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,4,power_law_1.01,1.1153663635253905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,4,power_law_1.01,1.4135040283203124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,4,power_law_1.01,2.1133567810058596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,4,balanced,1.5716479619344075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,4,power_law_1.01,4.31962890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,4,balanced,1.9650665918986003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,4,balanced,2.428778648376465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,4,power_law_1.01,0.06570879817008972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,4,power_law_1.01,0.0728384017944336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,4,power_law_1.01,0.08428800106048584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,4,power_law_1.01,0.07071359753608704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,4,power_law_1.01,0.08368639945983887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,4,power_law_1.01,0.0956928014755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,4,power_law_1.01,0.10315519571304321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,4,power_law_1.01,0.11582720279693604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,4,power_law_1.01,0.11413760185241699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,4,power_law_1.01,0.11909760236740112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,4,power_law_1.01,0.12043520212173461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,4,power_law_1.01,0.12556159496307373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,4,power_law_1.01,0.12893439531326295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,4,power_law_1.01,0.13246079683303832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,4,power_law_1.01,0.14261120557785034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,4,power_law_1.01,0.14328320026397706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,4,power_law_1.01,0.1561792016029358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,4,power_law_1.01,0.16983040571212768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,4,balanced,3.8209972381591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,4,power_law_1.01,0.19288320541381837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,4,power_law_1.01,0.2291327953338623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,4,power_law_1.01,0.2743936061859131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,4,power_law_1.01,0.34478719234466554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,4,power_law_1.01,0.4181375980377197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,4,power_law_1.01,0.5809472084045411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,4,power_law_1.01,0.7668992042541504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,4,power_law_1.01,1.0591808319091798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,4,power_law_1.01,1.4070783615112306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,4,balanced,4.743626594543457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,4,power_law_1.01,1.6559808731079102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,4,power_law_1.01,2.599500846862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,4,power_law_1.01,5.421420669555664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,4,power_law_1.2,0.07047680020332336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,4,power_law_1.2,0.07701759934425353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,4,power_law_1.2,0.07703040242195129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,4,power_law_1.2,0.08264319896697998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,4,power_law_1.2,0.09127039909362793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,4,power_law_1.2,0.10020480155944825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,4,power_law_1.2,0.106387197971344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,4,power_law_1.2,0.121670401096344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,4,power_law_1.2,0.12195839881896972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,4,power_law_1.2,0.12364799976348877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,4,power_law_1.2,0.12769919633865356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,4,power_law_1.2,0.12947200536727904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,4,power_law_1.2,0.13464959859848022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,4,power_law_1.2,0.13686399459838866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,4,power_law_1.2,0.14620800018310548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,4,power_law_1.2,0.14791040420532225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,4,power_law_1.2,0.15697920322418213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,4,power_law_1.2,0.17501440048217773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,4,power_law_1.2,0.18700799942016602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,4,power_law_1.2,0.2275264024734497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,4,power_law_1.2,0.25966720581054686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,8,balanced,0.037776000797748566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,8,balanced,0.03623999903599421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,8,balanced,0.03862933317820231
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,4,power_law_1.2,0.32343039512634275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,8,balanced,0.03510399907827377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,8,balanced,0.03914133210976919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,8,balanced,0.03639466563860575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,8,balanced,0.04074133435885111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,4,power_law_1.2,0.3882688045501709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,8,balanced,0.03793066740036011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,8,balanced,0.04112533231576284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,8,balanced,0.03903999924659729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,4,power_law_1.2,0.5443071842193603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,8,balanced,0.04041066765785217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,8,balanced,0.039077334105968475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,8,balanced,0.04113066693147024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,4,power_law_1.2,0.6811135768890381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,8,balanced,0.03868799904982249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,8,balanced,0.04249600072701772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,8,balanced,0.04062933226426443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,8,balanced,0.04028266668319702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,4,power_law_1.2,0.9413887977600097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,8,balanced,0.03899200012286504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,8,balanced,0.04065066576004028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,8,balanced,0.03867200016975403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,8,balanced,0.04061333338419596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,4,power_law_1.2,1.2477696418762207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,8,balanced,0.040991999208927155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,8,balanced,0.04102399945259094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,8,balanced,0.03908800085385641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,4,power_law_1.2,1.5982080459594727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,8,balanced,0.04252266883850098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,8,balanced,0.041402667760849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,8,balanced,0.040965333580970764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,8,balanced,0.04075733323891958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,4,power_law_1.2,2.342355155944824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,8,balanced,0.044677332043647766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,8,balanced,0.0447573314110438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,8,balanced,0.04688533147176107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,8,balanced,0.04470933477083842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,4,power_law_1.2,0.06530560255050659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,4,power_law_1.2,4.210367965698242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,8,balanced,0.04606399933497111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,8,balanced,0.04734399914741516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,8,balanced,0.048581331968307495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,4,power_law_1.2,0.07256960272789001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,8,balanced,0.04886400202910105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,8,balanced,0.0488319993019104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,4,power_law_1.2,0.06791039705276489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,8,balanced,0.05130666494369507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,8,balanced,0.053413331508636475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,8,balanced,0.058373332023620605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,4,power_law_1.2,0.06991360187530518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,8,balanced,0.05560533205668131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,8,balanced,0.06132799883683523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,4,power_law_1.2,0.08332160115242004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,8,balanced,0.06102400024731954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,4,power_law_1.2,0.09111679792404175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,8,balanced,0.07698666552702586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,4,power_law_1.2,0.1005247950553894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,8,balanced,0.06737599770228068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,8,balanced,0.09518933296203613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,4,power_law_1.2,0.11325440406799317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,8,balanced,0.08240533371766408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,8,balanced,0.1360266705354055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,4,power_law_1.2,0.11319040060043335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,4,power_law_1.2,0.11578240394592285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,8,balanced,0.16459199786186218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,8,balanced,0.09656000137329102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,4,power_law_1.2,0.11792000532150268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,4,power_law_1.2,0.1249343991279602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,4,power_law_1.2,0.12949119806289672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,8,balanced,0.23037334283192953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,8,balanced,0.13075199723243713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,4,power_law_1.2,0.13299200534820557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,4,power_law_1.2,0.14291839599609374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,8,balanced,0.2878506580988566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,4,power_law_1.2,0.1446720004081726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,8,balanced,0.15324800213178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,4,power_law_1.2,0.15459840297698973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,4,power_law_1.2,0.17592320442199708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,4,power_law_1.2,0.19359359741210938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,8,balanced,0.34461331367492676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,4,power_law_1.2,0.23402879238128663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,8,balanced,0.17730667193730673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,4,power_law_1.2,0.27486720085144045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,4,power_law_1.2,0.3617664098739624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,4,power_law_1.2,0.446457576751709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,8,balanced,0.5242720047632853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,4,power_law_1.2,0.61976318359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,4,power_law_1.2,0.7575808048248291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,8,balanced,0.2627093394597371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,4,power_law_1.2,1.1229632377624512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,4,power_law_1.2,1.42293119430542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,4,power_law_1.2,1.8100608825683593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,4,power_law_1.2,2.7426431655883787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,8,balanced,1.0044586658477783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,4,power_law_1.2,5.660851287841797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,8,balanced,0.48298664887746173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,8,power_law_1.01,0.03546879887580871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,8,power_law_1.01,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,8,power_law_1.01,0.03740800023078918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,8,power_law_1.01,0.038764798641204835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,8,power_law_1.01,0.039673599600791934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,8,power_law_1.01,0.037990400195121767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,8,power_law_1.01,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,8,power_law_1.01,0.04028159976005554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,8,power_law_1.01,0.041510400176048276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,8,power_law_1.01,0.041388800740242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,8,power_law_1.01,0.041945600509643556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,8,power_law_1.01,0.042559999227523806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,8,power_law_1.01,0.041254401206970215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,8,power_law_1.01,0.04213759899139404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,8,power_law_1.01,0.04567680060863495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,8,power_law_1.01,0.047788798809051514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,8,power_law_1.01,0.0486272007226944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,8,power_law_1.01,0.052070397138595584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,8,power_law_1.01,0.055852800607681274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,8,power_law_1.01,0.06359040141105651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,8,power_law_1.01,0.0711296021938324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,8,power_law_1.01,0.0880128026008606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,8,power_law_1.01,0.10662399530410767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,8,power_law_1.01,0.1356160044670105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,8,power_law_1.01,0.182425594329834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,8,power_law_1.01,0.26098558902740476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,8,power_law_1.01,0.3214848041534424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,8,power_law_1.01,0.4041152000427246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,8,power_law_1.01,0.5898816108703613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,8,power_law_1.01,1.1276415824890136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,8,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,8,power_law_1.2,0.03596160113811493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,8,power_law_1.01,0.034246399998664856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,8,power_law_1.2,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,8,power_law_1.2,0.038438400626182555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,8,power_law_1.01,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,8,power_law_1.2,0.03285120129585266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,8,power_law_1.2,0.036422398686408994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,8,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,8,power_law_1.2,0.03441919982433319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,8,power_law_1.2,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,8,power_law_1.01,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,8,power_law_1.2,0.03562879860401154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,8,power_law_1.2,0.03976320028305054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,8,power_law_1.01,0.03723520040512085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,8,power_law_1.2,0.03639039993286133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,8,power_law_1.2,0.039110401272773744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,8,power_law_1.01,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,8,power_law_1.2,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,8,power_law_1.01,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,8,power_law_1.2,0.039552000164985654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,8,power_law_1.01,0.03743360042572021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,8,power_law_1.2,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,8,power_law_1.2,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,8,power_law_1.01,0.03823359906673431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,8,power_law_1.2,0.03971840143203735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,8,power_law_1.2,0.038022398948669434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,8,power_law_1.01,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,8,power_law_1.2,0.040057599544525146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,8,power_law_1.2,0.03840000033378601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,8,power_law_1.01,0.03873920142650604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,8,power_law_1.2,0.04044159948825836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,8,power_law_1.2,0.03871999979019165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,8,power_law_1.01,0.04072319865226746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,8,power_law_1.2,0.04081920087337494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,8,power_law_1.2,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,8,power_law_1.01,0.03902080059051514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,8,power_law_1.2,0.04133760035037994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,8,power_law_1.2,0.03877759873867035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,8,power_law_1.01,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,8,power_law_1.2,0.04134399890899658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,8,power_law_1.2,0.040601599216461184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,8,power_law_1.01,0.04650239944458008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,8,power_law_1.2,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,8,power_law_1.2,0.04028800129890442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,8,power_law_1.01,0.04837760031223297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,8,power_law_1.2,0.046265599131584165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,8,power_law_1.2,0.04537599980831146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,8,power_law_1.01,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,8,power_law_1.2,0.04726400077342987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,8,power_law_1.2,0.04702720046043396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,8,power_law_1.01,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,8,power_law_1.2,0.05118719935417175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,8,power_law_1.2,0.04947839975357056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,8,power_law_1.01,0.06712960004806519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,8,power_law_1.2,0.05236480236053467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,8,power_law_1.01,0.07985280156135559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,8,power_law_1.2,0.05909759998321533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,8,power_law_1.01,0.11541119813919068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,8,power_law_1.2,0.06874240040779114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,8,power_law_1.01,0.14055039882659912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,8,power_law_1.2,0.0791167974472046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,8,power_law_1.01,0.19615999460220337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,8,power_law_1.2,0.11356799602508545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,8,power_law_1.01,0.241759991645813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,8,power_law_1.2,0.1366528034210205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,8,power_law_1.01,0.3454655885696411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,8,power_law_1.2,0.19171839952468872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,8,power_law_1.01,0.4266047954559326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,8,power_law_1.2,0.24274559020996095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,8,power_law_1.01,0.561356782913208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,8,power_law_1.2,0.35237119197845457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,8,power_law_1.01,0.8629695892333984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,8,power_law_1.2,0.45380477905273436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,8,power_law_1.2,0.5618624210357666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,8,power_law_1.01,1.7003904342651368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,8,power_law_1.2,0.874073600769043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,8,power_law_1.2,1.702681541442871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,8,balanced,0.044122666120529175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,8,balanced,0.0460746685663859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,8,balanced,0.04533333579699198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,8,balanced,0.047151997685432434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,8,balanced,0.04728533327579498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,8,balanced,0.04750399788220724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,8,balanced,0.04776533444722494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,8,balanced,0.048810665806134544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,8,balanced,0.04610133171081543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,8,balanced,0.04702933132648468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,8,balanced,0.04763199885686239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,8,balanced,0.0469706654548645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,8,balanced,0.04914666712284088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,8,balanced,0.04695466657479604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,8,balanced,0.05063466727733612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,8,balanced,0.05304533243179321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,8,balanced,0.052517334620157875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,8,balanced,0.05527999997138977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,8,balanced,0.054986665646235146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,8,balanced,0.06340266764163971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,8,balanced,0.06670933465162913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,8,balanced,0.07356800138950348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,8,balanced,0.08802666266759236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,8,balanced,0.10613333185513814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,8,balanced,0.12362666924794515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,8,balanced,0.16909867525100708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,8,balanced,0.2033066749572754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,8,balanced,0.25115732351938885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,8,balanced,0.03786666691303253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,8,balanced,0.038805333276589714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,8,balanced,0.0379573330283165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,8,balanced,0.04087999959786733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,8,balanced,0.03886933376391729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,8,balanced,0.04125866790612539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,8,balanced,0.03932266682386398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,8,balanced,0.04036800066630045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,8,balanced,0.04089066634575526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,8,balanced,0.36634135246276855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,8,balanced,0.041696002086003624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,8,balanced,0.041706666350364685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,8,balanced,0.04314666489760081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,8,balanced,0.04101866732041041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,8,balanced,0.04279999931653341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,8,balanced,0.04219200213750204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,8,balanced,0.04137066751718521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,8,balanced,0.04131733377774557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,8,balanced,0.042506664991378784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,8,balanced,0.04114133367935816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,8,balanced,0.04277333120505015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,8,balanced,0.042821332812309265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,8,balanced,0.04444266855716705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,8,balanced,0.042954668402671814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,8,balanced,0.04330133398373922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,8,balanced,0.04284266630808512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,8,balanced,0.044879997769991554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,8,balanced,0.04337066908677419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,8,balanced,0.050810664892196655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,8,balanced,0.04264533519744873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,8,balanced,0.04900800188382467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,8,balanced,0.048384000857671104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,8,balanced,0.05109866460164388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,8,balanced,0.6679626305898031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,8,balanced,0.04747733473777771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,8,balanced,0.0553653339544932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,8,balanced,0.049098665515581764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,8,balanced,0.05933333436648051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,8,balanced,0.05312533179918925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,8,balanced,0.06572266419728597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,8,balanced,0.057536001006762184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,8,balanced,0.0820853312810262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,8,balanced,0.059664001067479454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,8,balanced,0.11003200213114421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,8,balanced,0.0675786683956782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,8,balanced,0.13596266508102417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,8,balanced,0.08386666576067607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,8,balanced,0.175818661848704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,8,balanced,0.09847467144330342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,8,balanced,0.21884800990422568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,8,balanced,0.12432533502578735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,8,balanced,0.15130133430163065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,8,balanced,0.30903466542561847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,8,balanced,0.20057066281636557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,8,balanced,0.3882346550623576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,8,power_law_1.2,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,8,power_law_1.2,0.06378880143165588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,8,power_law_1.2,0.06855679750442505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,8,power_law_1.2,0.10014079809188843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,8,power_law_1.2,0.11203199625015259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,8,balanced,0.2546773354212443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,8,balanced,0.4782133499781291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,8,power_law_1.2,0.1418303966522217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,8,power_law_1.2,0.17221759557723998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,8,power_law_1.2,0.2480639934539795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,8,power_law_1.2,0.3274496078491211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,8,power_law_1.2,0.4052800178527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,8,power_law_1.2,0.6025152206420898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,8,balanced,0.7296000321706136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,8,power_law_1.2,1.052070426940918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,8,balanced,1.4075306256612141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,8,balanced,0.020362666497627895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,8,balanced,0.022101332743962605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,8,balanced,0.020917333662509918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,8,balanced,0.021573332448800404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,8,balanced,0.022005334496498108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,8,balanced,0.024656000236670177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,8,balanced,0.04708800216515859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,8,balanced,0.04722133278846741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,8,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,8,balanced,0.03301866600910822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,8,balanced,0.035216001172860466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,8,balanced,0.036703998843828835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,8,balanced,0.03623466690381368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,8,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,8,balanced,0.03687999894221624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,8,balanced,0.03865066667397817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,8,balanced,0.041146665811538696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,8,balanced,0.04400533437728882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,8,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,8,balanced,0.05514133473237356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,8,balanced,0.062261333068211876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,8,balanced,0.08666132887204488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,8,balanced,0.09967999656995137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,8,balanced,0.13541332880655924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,8,balanced,0.16845333576202393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,8,power_law_1.01,0.04307839870452881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,8,balanced,0.23822933435440063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,8,power_law_1.01,0.044409599900245664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,8,power_law_1.01,0.04305280148983002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,8,power_law_1.01,0.04503040015697479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,8,power_law_1.01,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,8,power_law_1.01,0.04509440064430237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,8,balanced,0.297050674756368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,8,power_law_1.01,0.047040000557899475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,8,power_law_1.01,0.04615679979324341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,8,power_law_1.01,0.03594239950180054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,8,power_law_1.01,0.0458624005317688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,8,power_law_1.01,0.03820799887180328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,8,power_law_1.01,0.04523519873619079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,8,power_law_1.01,0.03749760091304779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,8,power_law_1.01,0.04519039988517761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,8,balanced,0.3676746686299642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,8,power_law_1.01,0.039001598954200745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,8,power_law_1.01,0.046412798762321475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,8,power_law_1.01,0.04017919898033142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,8,power_law_1.01,0.04673280119895935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,8,power_law_1.01,0.04080640077590943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,8,power_law_1.01,0.04802559912204742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,8,power_law_1.01,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,8,power_law_1.01,0.05185920000076294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,8,power_law_1.01,0.039552000164985654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,8,power_law_1.01,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,8,power_law_1.01,0.054841601848602296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,8,balanced,0.5664533376693726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,8,power_law_1.01,0.06047999858856201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,8,power_law_1.01,0.06758400201797485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,8,power_law_1.01,0.07842559814453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,8,power_law_1.01,0.09230719804763794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,8,power_law_1.01,0.11370879411697388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,8,power_law_1.01,0.14378880262374877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,8,power_law_1.01,0.1943743944168091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,8,power_law_1.01,0.24416000843048097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,8,power_law_1.01,0.3632832050323486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,8,power_law_1.01,0.44301438331604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,8,power_law_1.01,0.5484672069549561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,8,power_law_1.01,0.8505663871765137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,8,balanced,1.0922613143920898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,8,power_law_1.01,1.6530176162719727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,8,balanced,0.3075786630312602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,8,balanced,0.45580800374348956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,8,balanced,0.8675146897633871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,8,power_law_1.01,0.03646720051765442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,8,power_law_1.01,0.03765760064125061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,8,power_law_1.01,0.03727999925613403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,8,power_law_1.01,0.03877759873867035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,8,power_law_1.01,0.03935999870300293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,8,power_law_1.01,0.04010240137577057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,8,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,8,power_law_1.01,0.041075199842453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,8,power_law_1.01,0.01950719952583313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,8,power_law_1.01,0.04168320000171662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,8,power_law_1.01,0.020268799364566804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,8,power_law_1.01,0.041440001130104064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,8,power_law_1.01,0.02080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,8,power_law_1.01,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,8,power_law_1.01,0.018828800320625304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,8,power_law_1.01,0.04322560131549835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,8,power_law_1.01,0.020703999698162077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,8,power_law_1.01,0.04337919950485229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,8,power_law_1.01,0.022944000363349915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,8,power_law_1.01,0.04425599873065948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,8,power_law_1.01,0.0452672004699707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,8,power_law_1.01,0.04712960124015808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,8,power_law_1.01,0.045228800177574156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,8,power_law_1.01,0.04748800098896026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,8,power_law_1.01,0.03430399894714355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,8,power_law_1.01,0.05053439736366272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,8,power_law_1.01,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,8,power_law_1.01,0.05578240156173706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,8,power_law_1.01,0.03348479866981506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.01,0.062745600938797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,8,power_law_1.01,0.033632001280784606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.01,0.0778432011604309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,8,power_law_1.01,0.03478400111198425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,8,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,8,power_law_1.01,0.03637759983539581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,8,power_law_1.01,0.03739520013332367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,8,power_law_1.01,0.04009599983692169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,8,power_law_1.01,0.04362240135669708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.01,0.0474368005990982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.01,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.01,0.06288639903068542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.01,0.08634880185127258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.01,0.0988864004611969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.01,0.13501440286636351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.01,0.16940799951553345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.01,0.23737599849700927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.01,0.29684479236602784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,8,power_law_1.01,0.3667072057723999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,8,power_law_1.01,0.5679743766784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,8,power_law_1.01,1.0897088050842285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,8,power_law_1.01,0.040934398770332336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,8,power_law_1.01,0.04226559996604919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,8,power_law_1.01,0.042630401253700254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,8,power_law_1.2,0.041920000314712526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,8,power_law_1.01,0.042931199073791504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,8,power_law_1.2,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,8,power_law_1.01,0.04391680061817169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,8,power_law_1.2,0.04270080029964447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,8,power_law_1.01,0.044896000623703004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,8,power_law_1.2,0.04503679871559143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,8,power_law_1.01,0.04938879907131195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,8,power_law_1.2,0.04520959854125976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,8,power_law_1.01,0.05119360089302063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,8,power_law_1.01,0.0543936014175415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,8,power_law_1.01,0.05964159965515137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,8,power_law_1.01,0.06718720197677612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,8,power_law_1.01,0.0850816011428833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,8,power_law_1.01,0.10982400178909302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,8,power_law_1.01,0.14563839435577391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,8,power_law_1.01,0.17973120212554933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,8,power_law_1.01,0.250547194480896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,8,power_law_1.01,0.31562879085540774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,8,power_law_1.01,0.45125761032104494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,8,power_law_1.01,0.5681280136108399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,8,power_law_1.01,0.7454271793365479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,8,power_law_1.01,1.1356800079345704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,8,power_law_1.01,2.193849563598633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,8,power_law_1.2,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,8,power_law_1.2,0.0383679986000061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,8,power_law_1.2,0.037785598635673524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,8,power_law_1.2,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,8,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,8,power_law_1.2,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,8,power_law_1.2,0.0404992014169693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,8,power_law_1.2,0.04100480079650879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,8,power_law_1.2,0.04225279986858368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,8,power_law_1.2,0.04161919951438904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,8,power_law_1.2,0.041580799221992495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,8,power_law_1.2,0.043942400813102724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,8,power_law_1.2,0.04350079894065857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,8,power_law_1.2,0.045049598813056944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,8,power_law_1.2,0.048979198932647704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,8,power_law_1.2,0.051801598072052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,8,power_law_1.2,0.05443199872970581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,8,power_law_1.2,0.0608959972858429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,8,power_law_1.2,0.06878719925880432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,8,power_law_1.2,0.08684800267219543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,8,power_law_1.2,0.10829440355300904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,8,power_law_1.2,0.14031360149383545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,8,power_law_1.2,0.17626880407333373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,8,power_law_1.2,0.24747519493103026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,8,power_law_1.2,0.32398080825805664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,8,power_law_1.2,0.035692799091339114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,8,power_law_1.2,0.4655360221862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,8,power_law_1.2,0.03639039993286133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,8,power_law_1.2,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,8,power_law_1.2,0.5908480167388916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,8,power_law_1.2,0.0377920001745224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,8,power_law_1.2,0.7449984073638916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,8,power_law_1.2,0.03976320028305054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,8,power_law_1.2,1.1567232131958007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,8,power_law_1.2,0.03961600065231323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,8,power_law_1.2,2.2668352127075195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,8,power_law_1.2,0.040038400888442995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,8,power_law_1.2,0.04031359851360321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,8,power_law_1.2,0.04043520092964172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,8,power_law_1.2,0.04082559943199158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,8,power_law_1.2,0.040940800309181215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,8,power_law_1.2,0.04343680143356323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.01,0.08750079870223999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,8,power_law_1.2,0.043084800243377686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.01,0.10931839942932128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,8,power_law_1.2,0.043699198961257936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.01,0.13359999656677246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,8,power_law_1.2,0.048876801133155824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.01,0.19463679790496827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,8,power_law_1.2,0.05009920001029968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.01,0.23712000846862794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,8,power_law_1.2,0.05146239995956421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.01,0.3441215991973877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,8,power_law_1.2,0.058278399705886844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.01,0.4314752101898193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.2,0.06399359703063964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.2,0.07964159846305847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,8,power_law_1.01,0.5346687793731689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.2,0.08924800157546997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,8,power_law_1.01,0.8153087615966796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.2,0.11481599807739258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,8,power_law_1.01,1.5727680206298829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.2,0.14168959856033325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.2,0.19046399593353272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.2,0.24238080978393556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.2,0.348140811920166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.2,0.45024638175964354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,8,power_law_1.2,0.5374335765838623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,8,power_law_1.2,0.8509951591491699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,8,power_law_1.2,1.6538496017456055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,8,power_law_1.2,0.0456063985824585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,8,power_law_1.2,0.04542720019817352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,8,power_law_1.2,0.04650880098342895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,8,power_law_1.2,0.04459519982337952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,8,power_law_1.2,0.04509440064430237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,8,power_law_1.2,0.04591360092163086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,8,power_law_1.2,0.04681600034236908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,8,power_law_1.2,0.04720639884471893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,8,power_law_1.2,0.04709759950637817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,8,power_law_1.2,0.05146239995956421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,8,power_law_1.2,0.05243520140647888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,8,power_law_1.2,0.0572160005569458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,8,power_law_1.2,0.05931519865989685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,8,power_law_1.2,0.06646400094032287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,8,power_law_1.2,0.08004480004310607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,8,power_law_1.2,0.09031680226325989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,8,power_law_1.2,0.12602239847183228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,8,power_law_1.2,0.14373120069503784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,8,power_law_1.2,0.19781119823455812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,8,power_law_1.2,0.2568320035934448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,8,power_law_1.2,0.3499648094177246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,8,power_law_1.2,0.45934720039367677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,8,power_law_1.2,0.5541567802429199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,8,power_law_1.2,0.8824895858764649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,8,power_law_1.2,1.7070592880249023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,8,power_law_1.2,0.019206400215625762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,8,power_law_1.2,0.02011519968509674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,8,power_law_1.2,0.02028159946203232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,8,power_law_1.2,0.019411200284957887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,8,power_law_1.2,0.020428800582885744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,8,power_law_1.2,0.022227199375629426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,8,power_law_1.2,0.04552960097789764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,8,power_law_1.2,0.045484799146652224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,8,power_law_1.2,0.03209599852561951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,8,power_law_1.2,0.03219839930534363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,8,power_law_1.2,0.03315199911594391
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,8,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,8,power_law_1.2,0.03429119884967804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,8,power_law_1.2,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,8,power_law_1.2,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,8,power_law_1.2,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,8,power_law_1.2,0.04026240110397339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,8,power_law_1.2,0.04302079975605011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.2,0.04648320078849792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.2,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.2,0.060755199193954466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.2,0.08633599877357483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.2,0.0986303985118866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.2,0.13372160196304322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.2,0.16897280216217042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.2,0.2375040054321289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.2,0.2975167989730835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,8,power_law_1.2,0.3671999931335449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,8,power_law_1.2,0.5682496070861817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,8,power_law_1.2,1.0932160377502442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,256,balanced,0.08879466851552327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,256,balanced,0.0853760043780009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,256,balanced,0.08823466300964355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,256,balanced,0.08556800087292989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,256,balanced,0.09340799848238628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,256,balanced,0.09114666779836018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,256,balanced,0.09241066376368205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,256,balanced,0.09406399726867676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,256,balanced,0.06317866841952006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,256,balanced,0.09609066446622212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,256,balanced,0.062447999914487205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,256,balanced,0.09470933675765991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,256,balanced,0.06506666541099548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,256,balanced,0.09487467010815938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,256,balanced,0.06666133304437001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,256,balanced,0.06490133206049602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,256,balanced,0.09312533338864644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,256,balanced,0.07010666529337566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,256,balanced,0.09473599990208943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,256,balanced,0.069882666071256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,256,balanced,0.09171733260154724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,256,balanced,0.06896533568700154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,256,balanced,0.10046933094660442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,256,balanced,0.06995200117429097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,256,balanced,0.1002400020758311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,256,balanced,0.07156800230344136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,256,balanced,0.06911999980608623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,256,balanced,0.10114666819572449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,256,balanced,0.07313600182533264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,256,balanced,0.10442133744557698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,256,balanced,0.07539199789365132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,256,balanced,0.07349333167076111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,256,balanced,0.10778666536013286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,256,balanced,0.08049599826335907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,256,balanced,0.08186133205890656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,256,balanced,0.11414933204650879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,256,balanced,0.08347733815511067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,256,balanced,0.08929600318272908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,256,balanced,0.12257066369056702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,256,balanced,0.09596266349156697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,256,balanced,0.1432213286558787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,256,balanced,0.11587733030319214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,256,balanced,0.13688000043233237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,256,balanced,0.16154133280118307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,256,balanced,0.17614400386810303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,256,balanced,0.20121065775553384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,256,balanced,0.20813866456349692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,256,balanced,0.2754240036010742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,256,balanced,0.06095466514428457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,256,balanced,0.23874666293462118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,256,balanced,0.061568001906077065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,256,balanced,0.06043200194835663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,256,balanced,0.06422933439413707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,256,balanced,0.06755200028419495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,256,balanced,0.06529066463311513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,256,balanced,0.06765866776307423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,256,balanced,0.34195733070373535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,256,balanced,0.06957866748174031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,256,balanced,0.06901866694291432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,256,balanced,0.06730133295059204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,256,balanced,0.3187573353449504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,256,balanced,0.06975999971230824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,256,balanced,0.06930666665236156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,256,balanced,0.06955199937025706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,256,balanced,0.07115733126799266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,256,balanced,0.075013334552447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,256,balanced,0.47953065236409503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,256,balanced,0.07729066908359528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,256,balanced,0.07957333326339722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,256,balanced,0.08533866206804912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,256,balanced,0.08967999617258708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,256,balanced,0.39265068372090656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,256,balanced,0.1018453339735667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,256,balanced,0.11189333597819011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,256,balanced,0.613482673962911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,256,balanced,0.13849600156148276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,256,balanced,0.16507200400034586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,256,balanced,0.484554648399353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,256,balanced,0.21456533670425415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,256,balanced,0.7528693675994873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,256,balanced,0.2579360008239746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,256,balanced,0.023311999936898548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,256,balanced,0.3590506712595622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,256,balanced,0.02409599969784419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,256,balanced,0.024906667570273083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,256,balanced,0.028832000990708668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,256,balanced,0.0533493310213089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,256,balanced,0.7124000390370687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,256,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,256,balanced,0.04474666714668274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,256,balanced,0.043925335009892784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,256,balanced,0.0446720023949941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,256,balanced,1.1498026847839355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,256,balanced,0.04520000020662943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,256,balanced,0.045050665736198425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,256,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,256,balanced,0.04281599819660187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,256,balanced,0.44832531611124676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,256,balanced,0.04244266450405121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,256,balanced,0.05053333441416422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,256,balanced,0.049322664737701416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,256,balanced,0.04692799846331278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,256,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,256,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,256,balanced,0.08077333370844524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,256,balanced,0.08813333511352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,256,balanced,0.546175996462504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,256,balanced,0.12211733063062032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,256,balanced,0.15132799744606018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,256,balanced,0.21414399147033691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,256,balanced,1.3930773735046387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,256,balanced,0.2746293346087138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,256,balanced,2.263381322224935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,256,balanced,0.832757314046224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,256,balanced,0.397866686185201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,256,balanced,0.5219253301620483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,256,balanced,0.6435093482335409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,256,balanced,1.6220426559448242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,256,balanced,1.0102933247884114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,256,power_law_1.01,0.06104959845542908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,256,power_law_1.01,0.07894399762153625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,256,power_law_1.01,0.06229760050773621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,256,power_law_1.01,0.08107519745826722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,256,power_law_1.01,0.06202239990234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,256,power_law_1.01,0.07967360019683838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,256,power_law_1.01,0.0806656002998352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,256,power_law_1.01,0.0656000018119812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,256,power_law_1.01,0.08355839848518372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,256,power_law_1.01,0.06563199758529663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,256,power_law_1.01,0.0862335979938507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,256,power_law_1.01,0.06804479956626892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,256,power_law_1.01,0.08680959939956664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,256,power_law_1.01,0.0685375988483429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,256,power_law_1.01,0.08389760255813598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,256,power_law_1.01,0.0670144021511078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,256,power_law_1.01,0.08833280205726624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,256,power_law_1.01,0.06696959733963012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,256,power_law_1.01,0.08509439826011658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,256,power_law_1.01,0.06848639845848084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,256,power_law_1.01,0.08332160115242004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,256,power_law_1.01,0.069651198387146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,256,power_law_1.01,0.08488960266113281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,256,power_law_1.01,0.06839039921760559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,256,power_law_1.01,0.08240640163421631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,256,power_law_1.01,0.06892160177230836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,256,power_law_1.01,0.08186240196228027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,256,power_law_1.01,0.06912639737129211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,256,balanced,1.9814186096191406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,256,power_law_1.01,0.08485119938850402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,256,power_law_1.01,0.07155200242996215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,256,power_law_1.01,0.08326399922370911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,256,power_law_1.01,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,256,power_law_1.01,0.08406400084495544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,256,power_law_1.01,0.07405440211296081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,256,power_law_1.01,0.10384639501571655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,256,power_law_1.01,0.09084799885749817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,256,power_law_1.01,0.10533119440078735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,256,power_law_1.01,0.0944703996181488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,256,power_law_1.01,0.12142080068588257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,256,power_law_1.01,0.13027199506759643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,256,power_law_1.01,0.13802239894866944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,256,power_law_1.01,0.15733120441436768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,256,power_law_1.01,0.18751360177993776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,256,power_law_1.01,0.20082559585571289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,256,power_law_1.01,0.22693119049072266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,256,power_law_1.01,0.24010241031646729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,256,power_law_1.01,0.2889535903930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,256,power_law_1.01,0.3263488054275513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,256,power_law_1.01,0.39059200286865237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,256,power_law_1.01,0.4162879943847656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,256,power_law_1.01,0.5528319835662842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,256,power_law_1.01,0.6226624011993408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,256,power_law_1.01,0.6868607997894287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,256,power_law_1.01,0.7696767807006836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,256,power_law_1.01,0.8250240325927735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,256,power_law_1.01,0.9531328201293945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,256,power_law_1.01,1.3667008399963378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,256,power_law_1.01,1.5610367774963378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,256,power_law_1.01,2.7652544021606444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,256,power_law_1.01,3.034252738952637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,256,power_law_1.01,0.06080639958381653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,256,power_law_1.01,0.060147202014923094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,256,power_law_1.01,0.06472319960594178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,256,power_law_1.01,0.06343680024147033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,256,power_law_1.01,0.06403840184211732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,256,power_law_1.01,0.06440320014953613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,256,power_law_1.01,0.06757760047912598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,256,power_law_1.01,0.06590080261230469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,256,power_law_1.01,0.06548479795455933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,256,power_law_1.01,0.06632959842681885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,256,power_law_1.01,0.06652799844741822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,256,power_law_1.01,0.06652799844741822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,256,power_law_1.01,0.06787199974060058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,256,power_law_1.01,0.06573439836502075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,256,power_law_1.01,0.07174400091171265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,256,power_law_1.01,0.06900479793548583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,256,power_law_1.01,0.07071359753608704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,256,power_law_1.01,0.08687360286712646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.01,0.08797439932823181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.01,0.11070719957351685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.01,0.12509440183639525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.01,0.16424959897994995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.01,0.20034561157226563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.01,0.2657279968261719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.01,0.32385280132293703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.01,0.45705599784851075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.01,0.6122879981994629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,256,power_law_1.01,0.7565311908721923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,256,power_law_1.01,1.1803327560424806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,256,power_law_1.01,2.499705505371094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,256,power_law_1.01,0.021580800414085388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,256,power_law_1.01,0.021862399578094483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,256,power_law_1.01,0.02309119999408722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,256,power_law_1.01,0.02813439965248108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,256,power_law_1.01,0.05190399885177612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,256,power_law_1.01,0.04364160001277924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,256,power_law_1.01,0.042335999011993405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,256,power_law_1.01,0.043059200048446655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,256,power_law_1.01,0.04307839870452881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,256,power_law_1.01,0.043750399351119997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,256,power_law_1.01,0.04318720102310181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,256,power_law_1.01,0.043372800946235655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,256,power_law_1.01,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,256,power_law_1.01,0.04069119989871979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,256,power_law_1.01,0.04219520092010498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,256,power_law_1.01,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,256,power_law_1.01,0.0489984005689621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,256,power_law_1.01,0.050732797384262084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.01,0.05334399938583374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,256,power_law_1.2,0.061913597583770755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,256,power_law_1.2,0.08461440205574036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.01,0.07130240201950074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,256,power_law_1.2,0.06357120275497437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,256,power_law_1.2,0.07947520017623902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.01,0.0802623987197876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,256,power_law_1.2,0.06620799899101257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,256,power_law_1.2,0.08042880296707153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.01,0.10559999942779541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,256,power_law_1.2,0.06572160124778748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,256,power_law_1.2,0.08574720025062561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.01,0.12961280345916748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,256,power_law_1.2,0.06653439998626709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,256,power_law_1.2,0.08481280207633972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.01,0.18428800106048585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,256,power_law_1.2,0.06743680238723755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,256,power_law_1.2,0.08791040182113648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.01,0.2420799970626831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,256,power_law_1.2,0.06824960112571717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,256,power_law_1.2,0.08521599769592285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.01,0.3435391902923584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,256,power_law_1.2,0.0861952006816864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,256,power_law_1.2,0.06855040192604064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.01,0.44135680198669436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,256,power_law_1.2,0.08635519742965699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,256,power_law_1.2,0.06880000233650208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,256,power_law_1.01,0.5203775882720947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,256,power_law_1.2,0.08446080088615418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,256,power_law_1.2,0.06933760046958923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,256,power_law_1.01,0.8844608306884766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,256,power_law_1.2,0.08123520016670227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,256,power_law_1.2,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,256,power_law_1.2,0.08399360179901123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,256,power_law_1.2,0.06803200244903565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,256,power_law_1.01,1.7682687759399414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,256,power_law_1.2,0.08223999738693237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,256,power_law_1.2,0.06762239933013917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,256,power_law_1.2,0.07848320007324219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,256,power_law_1.2,0.06793599724769592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,256,power_law_1.2,0.08241919875144958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,256,power_law_1.2,0.0705344021320343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,256,power_law_1.2,0.07004799842834472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,256,power_law_1.2,0.08299520015716552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,256,power_law_1.2,0.08441600203514099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,256,power_law_1.2,0.09136639833450318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,256,power_law_1.2,0.10283520221710205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,256,power_law_1.2,0.08896639943122864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,256,power_law_1.2,0.11296639442443848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,256,power_law_1.2,0.10647679567337036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,256,power_law_1.2,0.14878720045089722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,256,power_law_1.2,0.1451647996902466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,256,power_law_1.2,0.16903040409088135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,256,power_law_1.2,0.16834559440612792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,256,power_law_1.2,0.22268800735473632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,256,power_law_1.2,0.22824320793151856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,256,power_law_1.2,0.28769280910491946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,256,power_law_1.2,0.26997759342193606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,256,power_law_1.2,0.3980479955673218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,256,power_law_1.2,0.43048319816589353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,256,power_law_1.2,0.5150720119476319
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,256,power_law_1.2,0.5768511772155762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,256,power_law_1.2,0.8196479797363281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,256,power_law_1.2,0.7723648071289062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,256,power_law_1.2,1.0370047569274903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,256,power_law_1.2,1.0187456130981445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,256,power_law_1.2,1.3318655967712403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,256,power_law_1.2,1.3003328323364258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,256,power_law_1.2,2.2456640243530273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,256,power_law_1.2,2.127238464355469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,256,power_law_1.2,0.05886719822883606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,256,power_law_1.2,4.801683044433593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,256,power_law_1.2,4.546873474121094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,256,power_law_1.2,0.05910400152206421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,256,power_law_1.2,0.0598143994808197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,256,power_law_1.2,0.06422399878501892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,256,power_law_1.2,0.06496000289916992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,256,power_law_1.2,0.06570240259170532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,256,power_law_1.2,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,256,power_law_1.2,0.06580479741096497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,256,power_law_1.2,0.06558079719543457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,256,power_law_1.2,0.06524800062179566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,256,power_law_1.2,0.06439039707183838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,256,power_law_1.2,0.0644864022731781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,256,power_law_1.2,0.06369280219078063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,256,power_law_1.2,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,256,power_law_1.2,0.0692351996898651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,256,power_law_1.2,0.06614400148391723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,256,power_law_1.2,0.08357120156288148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,256,power_law_1.2,0.08426240086555481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.2,0.0998080015182495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.2,0.12103040218353271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.2,0.14722559452056885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.2,0.19026559591293335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.2,0.22593278884887696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.2,0.31560320854187013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.2,0.40097918510437014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.2,0.6016128063201904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.2,0.8092991828918457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,256,power_law_1.2,1.0013567924499511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,256,power_law_1.2,1.7631296157836913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,256,power_law_1.2,3.4844608306884766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,256,power_law_1.2,0.021766400337219237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,256,power_law_1.2,0.022368000447750093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,256,power_law_1.2,0.02234880030155182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,256,power_law_1.2,0.027046400308609008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,256,power_law_1.2,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,256,power_law_1.2,0.04172160029411316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,256,power_law_1.2,0.04232319891452789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,256,power_law_1.2,0.042124798893928526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,256,power_law_1.2,0.04380159974098206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,256,power_law_1.2,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,256,power_law_1.2,0.04328320026397705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,256,power_law_1.2,0.042444801330566405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,256,power_law_1.2,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,256,power_law_1.2,0.039980798959732056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,256,power_law_1.2,0.042982399463653564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,128,balanced,0.05420266588528951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,128,balanced,0.06525333225727081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,128,balanced,0.05465066432952881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,256,power_law_1.2,0.046937599778175354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,128,balanced,0.06875733534495036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,128,balanced,0.055760001142819725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,256,power_law_1.2,0.04849919974803925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,128,balanced,0.06759466727574666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,128,balanced,0.05952000121275584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,128,balanced,0.07146133482456207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,128,balanced,0.05727999905745188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,256,power_law_1.2,0.050988799333572386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,128,balanced,0.07307200133800507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,128,balanced,0.0639573335647583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.2,0.06037120223045349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,128,balanced,0.07870399951934814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,128,balanced,0.06345599889755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,128,balanced,0.08162133395671844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,128,balanced,0.06436799963315327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.2,0.07192320227622986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,128,balanced,0.08061333497365315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,128,balanced,0.06469333171844482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.2,0.08638719916343689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,128,balanced,0.08348799745241801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,128,balanced,0.06755200028419495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,128,balanced,0.07899199922879536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.2,0.11066240072250366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,128,balanced,0.06713599960009257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,128,balanced,0.08201066652933757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,128,balanced,0.06533866624037425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.2,0.14064639806747437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,128,balanced,0.0846613347530365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,128,balanced,0.07310399909814198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.2,0.19078400135040283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,128,balanced,0.085999995470047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,128,balanced,0.0709440012772878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.2,0.24820480346679688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,128,balanced,0.07965866724650066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,128,balanced,0.07313066720962524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,128,balanced,0.08556266625722249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.2,0.3619584083557129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,128,balanced,0.07482666770617168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,128,balanced,0.09230400125185649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,128,balanced,0.07991999884446462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.2,0.4814271926879883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,128,balanced,0.09354666868845622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,128,balanced,0.08561600248018901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,256,power_law_1.2,0.5926527976989746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,128,balanced,0.09623466928799947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,128,balanced,0.09289600451787312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,256,power_law_1.2,0.9789888381958007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,128,balanced,0.10447466373443604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,128,balanced,0.11421333750089009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,256,power_law_1.2,1.8240959167480468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,128,balanced,0.10961066683133443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,128,balanced,0.13383466998736063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,128,balanced,0.12029866377512614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,128,balanced,0.17547200123469034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,128,balanced,0.14131733775138855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,128,balanced,0.20794665813446045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,128,balanced,0.15980799992879233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,128,balanced,0.2785759965578715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,128,balanced,0.05613866448402405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,128,balanced,0.20317333936691284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,128,balanced,0.05550933380921682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,128,balanced,0.056976000467936196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,128,balanced,0.3454666535059611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,128,balanced,0.05916800101598104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,128,balanced,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,128,balanced,0.06554666658242543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,128,balanced,0.06711466610431671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,128,balanced,0.0653599997361501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,128,balanced,0.24155733982721964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,128,balanced,0.06740266581376393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,128,balanced,0.0671253353357315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,128,balanced,0.06494933366775513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,128,balanced,0.06752533217271169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,128,balanced,0.06933866441249847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,128,balanced,0.4832106828689575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,128,balanced,0.0720960001150767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,128,balanced,0.07482666770617168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,128,balanced,0.07704000174999237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,128,balanced,0.32047466437021893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,128,balanced,0.07730666796366374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,128,balanced,0.08326399823029836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,128,balanced,0.09098666906356812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,128,balanced,0.1006666620572408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,128,balanced,0.6179626782735189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,128,balanced,0.10924800237019856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,128,balanced,0.39451201756795246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,128,balanced,0.14267200231552124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,128,balanced,0.16435733437538147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,128,balanced,0.7599573135375977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,128,balanced,0.21497066815694174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,128,balanced,0.49245333671569824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,128,balanced,0.2609600027402242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,128,balanced,0.3599840005238851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,128,balanced,1.1619253158569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,128,balanced,0.7203733126322428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,128,balanced,0.4570346673329671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,128,balanced,0.5532533327738444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,128,balanced,0.022592000663280487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,128,balanced,0.024133334557215374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,128,balanced,0.023919999599456787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,128,balanced,0.02571200082699458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,128,balanced,0.028565332293510437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,128,balanced,0.053541332483291626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,128,balanced,0.04470933477083842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,128,balanced,0.044266665975252785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,128,balanced,0.04484266539414724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,128,balanced,0.045328001181284584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,128,balanced,0.04460800190766653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,128,balanced,2.2872427304585776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,128,balanced,0.04381333291530609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,128,balanced,0.04308799902598063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,128,balanced,0.04253333310286204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,128,balanced,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,128,balanced,0.054341331124305725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,128,balanced,0.8432586987813314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,128,balanced,0.048528000712394714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,128,balanced,1.40666659673055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,128,balanced,0.052576000491778054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,128,balanced,0.0591839998960495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,128,balanced,0.08427733182907104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,128,balanced,0.09405333797136943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,128,balanced,0.1285599966843923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,128,balanced,0.1600160002708435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,128,balanced,0.22340265909830728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,128,balanced,0.2796586751937866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,128,balanced,1.6490933100382488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,128,balanced,0.4159253438313802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,128,balanced,0.5443040132522583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,128,balanced,0.671066681543986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,128,power_law_1.01,0.07038080096244811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,128,power_law_1.01,0.057036799192428586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,128,power_law_1.01,0.0706496000289917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,128,power_law_1.01,0.05509759783744812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,128,power_law_1.01,0.0713919997215271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,128,power_law_1.01,0.05533440113067627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,128,power_law_1.01,0.07192959785461425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,128,power_law_1.01,0.057529598474502563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,128,power_law_1.01,0.07461119890213012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,128,power_law_1.01,0.06080639958381653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,128,power_law_1.01,0.07555840015411378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,128,power_law_1.01,0.0606719970703125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,128,power_law_1.01,0.07743359804153442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,128,power_law_1.01,0.06334720253944397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,128,power_law_1.01,0.07521920204162598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,128,power_law_1.01,0.06094080209732056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,128,power_law_1.01,0.07728000283241272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,128,power_law_1.01,0.06371840238571166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,128,power_law_1.01,0.07701759934425353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,128,power_law_1.01,0.0636031985282898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,128,balanced,1.05404798189799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,128,power_law_1.01,0.06440320014953613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,128,power_law_1.01,0.07572479844093323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,128,power_law_1.01,0.0637503981590271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,128,power_law_1.01,0.07854719758033753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,128,power_law_1.01,0.06717439889907836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,128,power_law_1.01,0.07978879809379577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,128,power_law_1.01,0.06581760048866273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,128,power_law_1.01,0.080595201253891
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,128,power_law_1.01,0.08468480110168457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,128,power_law_1.01,0.07207679748535156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,128,power_law_1.01,0.08357120156288148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,128,power_law_1.01,0.07227519750595093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,128,power_law_1.01,0.07484800219535828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,128,power_law_1.01,0.08615040183067321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,128,power_law_1.01,0.0875328004360199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,128,power_law_1.01,0.09621760249137878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,128,power_law_1.01,0.12639360427856444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,128,power_law_1.01,0.14835840463638306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,128,power_law_1.01,0.19153280258178712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,128,power_law_1.01,0.235916805267334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,128,power_law_1.01,0.3173952102661133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,128,power_law_1.01,0.41050238609313966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,128,power_law_1.01,0.5810624122619629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,128,power_law_1.01,0.7366911888122558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,128,power_law_1.01,0.9271039962768555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,128,power_law_1.01,1.539302444458008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,128,power_law_1.01,2.8962751388549806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,128,power_law_1.01,0.06431360244750976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,128,power_law_1.01,0.0644927978515625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,128,power_law_1.01,0.058233600854873654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,128,power_law_1.01,0.058387202024459836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,128,power_law_1.01,0.06401919722557067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,128,power_law_1.01,0.061401599645614625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,128,power_law_1.01,0.06467199921607972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,128,power_law_1.01,0.06371200084686279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,128,power_law_1.01,0.06316800117492676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,128,power_law_1.01,0.06442239880561829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,128,power_law_1.01,0.06502400040626526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,128,power_law_1.01,0.06667519807815551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,128,power_law_1.01,0.06585599780082703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,128,power_law_1.01,0.06627200245857238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,128,power_law_1.01,0.07059199810028076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,128,power_law_1.01,0.07073919773101807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,128,power_law_1.01,0.07507839798927307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,128,power_law_1.01,0.08299520015716552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,128,balanced,2.0736586252848306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.01,0.09340800046920776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.01,0.11351679563522339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.01,0.1271615982055664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.01,0.16011519432067872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.01,0.19306880235671997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.01,0.254419207572937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.01,0.3211008071899414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.01,0.4534719944000244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.01,0.5689599990844727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,128,power_law_1.01,0.7543680191040039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,128,power_law_1.01,1.138771152496338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,128,power_law_1.01,2.1661888122558595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,128,power_law_1.01,0.020787200331687926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,128,power_law_1.01,0.020921599864959717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,128,power_law_1.01,0.021887999773025513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,128,power_law_1.01,0.023238399624824525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,128,power_law_1.01,0.027475199103355406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,128,power_law_1.01,0.0995136022567749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,128,power_law_1.01,0.050886398553848265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,128,power_law_1.01,0.10623999834060668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,128,power_law_1.01,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,128,power_law_1.01,0.11967359781265259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,128,power_law_1.01,0.04126720130443573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,128,power_law_1.01,0.13687039613723756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,128,power_law_1.01,0.04127359986305237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,128,power_law_1.01,0.1739776015281677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,128,power_law_1.01,0.04219520092010498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,128,power_law_1.01,0.2062079906463623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,128,power_law_1.01,0.04280959963798523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,128,power_law_1.01,0.270630407333374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,128,power_law_1.01,0.04335359930992126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,128,power_law_1.01,0.3418495893478394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,128,power_law_1.01,0.04246399998664856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,128,power_law_1.01,0.47945599555969237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,128,power_law_1.01,0.04203520119190216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,128,power_law_1.01,0.6192512035369873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,128,power_law_1.01,0.04180479943752289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,128,power_law_1.01,0.7342527866363525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,128,power_law_1.01,0.04240640103816986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,128,power_law_1.01,1.274662399291992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,128,power_law_1.01,0.048230400681495665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,128,power_law_1.01,0.049158400297164916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,128,power_law_1.01,2.220204734802246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.01,0.05320320129394531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,128,power_law_1.2,0.0711296021938324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.01,0.06639999747276307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,128,power_law_1.2,0.07542399764060974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.01,0.08016639947891235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,128,power_law_1.2,0.06997759938240052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.01,0.10495359897613525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,128,power_law_1.2,0.07356799840927124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.01,0.12252800464630127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,128,power_law_1.2,0.07434880137443542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.01,0.1831104040145874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,128,power_law_1.2,0.07568640112876893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.01,0.23047680854797364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,128,power_law_1.2,0.07690240144729614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.01,0.31112959384918215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,128,power_law_1.2,0.07820799946784973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.01,0.42162561416625977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,128,power_law_1.2,0.07648640275001525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,128,power_law_1.01,0.5108096122741699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,128,power_law_1.2,0.07749760150909424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,128,power_law_1.01,0.841107177734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,128,power_law_1.2,0.07889919877052307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,128,power_law_1.2,0.07873280048370361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,128,power_law_1.01,1.6561216354370116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,128,power_law_1.2,0.08247039914131164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,128,power_law_1.2,0.08068479895591736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,128,power_law_1.2,0.05813120007514953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,128,power_law_1.2,0.08612480163574218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,128,power_law_1.2,0.060889601707458496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,128,power_law_1.2,0.08632959723472595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,128,power_law_1.2,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,128,power_law_1.2,0.09545599818229675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,128,power_law_1.2,0.05921279788017273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,128,power_law_1.2,0.10321279764175414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,128,power_law_1.2,0.06047999858856201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,128,power_law_1.2,0.11081600189208984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,128,power_law_1.2,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,128,power_law_1.2,0.12482559680938721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,128,power_law_1.2,0.06343039870262146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,128,power_law_1.2,0.1574720025062561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,128,power_law_1.2,0.0642304003238678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,128,power_law_1.2,0.198854398727417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,128,power_law_1.2,0.06477439999580384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,128,power_law_1.2,0.23359360694885253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,128,power_law_1.2,0.06478719711303711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,128,power_law_1.2,0.3144320011138916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,128,power_law_1.2,0.06584320068359376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,128,power_law_1.2,0.06350719928741455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,128,power_law_1.2,0.3892287969589233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,128,power_law_1.2,0.06776319742202759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,128,power_law_1.2,0.6539648056030274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,128,power_law_1.2,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,128,power_law_1.2,0.804742431640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,128,power_law_1.2,0.07077119946479797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,128,power_law_1.2,1.0178560256958007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,128,power_law_1.2,0.0720255970954895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,128,power_law_1.2,1.7572160720825196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,128,power_law_1.2,0.0804095983505249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,128,power_law_1.2,0.09085440039634704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,128,power_law_1.2,3.5026302337646484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,128,power_law_1.2,0.10314240455627441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,128,power_law_1.2,0.13297280073165893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,128,power_law_1.2,0.15842560529708863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,128,power_law_1.2,0.20899200439453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,128,power_law_1.2,0.26808319091796873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,128,power_law_1.2,0.3706432104110718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,128,power_law_1.2,0.468339204788208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,128,power_law_1.2,0.6859583854675293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,128,power_law_1.2,0.8988096237182617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,128,power_law_1.2,1.1008319854736328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,128,power_law_1.2,1.7745664596557618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,128,power_law_1.2,4.027660751342774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,128,power_law_1.2,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,128,power_law_1.2,0.06536960005760192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,128,power_law_1.2,0.0568448007106781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,128,power_law_1.2,0.061312001943588254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,128,power_law_1.2,0.06338559985160827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,128,power_law_1.2,0.063372802734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,128,power_law_1.2,0.06357759833335877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,128,power_law_1.2,0.06424959897994995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,128,power_law_1.2,0.06419199705123901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,128,power_law_1.2,0.0642624020576477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,128,power_law_1.2,0.06653439998626709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,128,power_law_1.2,0.06569600105285645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,128,power_law_1.2,0.06599680185317994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,128,power_law_1.2,0.021164800226688384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,128,power_law_1.2,0.0664896011352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,128,power_law_1.2,0.021887999773025513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,128,power_law_1.2,0.07046399712562561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,128,power_law_1.2,0.02194560021162033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,128,power_law_1.2,0.07043840289115906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,128,power_law_1.2,0.022950400412082673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,128,power_law_1.2,0.07453439831733703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,128,power_law_1.2,0.0278656005859375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,128,power_law_1.2,0.08797439932823181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,128,power_law_1.2,0.05204480290412903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,128,power_law_1.2,0.041171199083328246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.2,0.09498239755630493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,128,power_law_1.2,0.04311679899692535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.2,0.11713919639587403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,128,power_law_1.2,0.04176000058650971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.2,0.13523199558258056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,128,power_law_1.2,0.04185599982738495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.2,0.17185280323028565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,128,power_law_1.2,0.04325119853019714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.2,0.2130176067352295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,128,power_law_1.2,0.04222080111503601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.2,0.30651519298553465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,128,power_law_1.2,0.04145280122756958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.2,0.36126079559326174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,128,power_law_1.2,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.2,0.5334144115447998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,128,power_law_1.2,0.04199039936065674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.2,0.7606272220611572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,128,power_law_1.2,0.04471679925918579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,128,power_law_1.2,0.9021632194519043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,128,power_law_1.2,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,128,power_law_1.2,1.5627967834472656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,128,power_law_1.2,0.049983999133110045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.2,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,128,power_law_1.2,3.126092720031738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.2,0.06986879706382751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.2,0.08640000224113464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.2,0.11058559417724609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,64,balanced,0.060133333007494606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.2,0.1329856038093567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,64,balanced,0.06022400160630544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.2,0.20392320156097413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,64,balanced,0.06121066709359487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,64,balanced,0.06347733239332835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.2,0.2309567928314209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,64,balanced,0.06636266907056172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.2,0.3645119905471802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,64,balanced,0.07484266658624013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,64,balanced,0.07773333291212718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.2,0.4698239803314209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,64,balanced,0.07906666894753774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,128,power_law_1.2,0.5698048114776612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,64,balanced,0.07734933495521545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,64,balanced,0.07745066781838734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,128,power_law_1.2,0.9086015701293946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,64,balanced,0.07854400078455608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,64,balanced,0.07951466739177704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,128,power_law_1.2,1.8812736511230468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,64,balanced,0.08330133557319641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,64,balanced,0.0802346666653951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,64,balanced,0.052282666166623436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,64,balanced,0.05120533208052317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,64,balanced,0.08564266562461853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,64,balanced,0.05355200171470642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,64,balanced,0.08941333492596944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,64,balanced,0.05559466779232025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,64,balanced,0.05689600110054016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,64,balanced,0.08910933136940002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,64,balanced,0.06346133351325989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,64,balanced,0.0627040018637975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,64,balanced,0.09607999523480733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,64,balanced,0.0642986645301183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,64,balanced,0.06727466483910878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,64,balanced,0.10014399886131287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,64,balanced,0.06623999774456024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,64,balanced,0.06818133095900218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,64,balanced,0.11291733384132385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,64,balanced,0.06613333523273468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,64,balanced,0.07123733560244243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,64,balanced,0.06884799897670746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,64,balanced,0.12300266822179158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,64,balanced,0.0740586668252945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,64,balanced,0.0768746683994929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,64,balanced,0.07977066437403361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,64,balanced,0.1460479994614919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,64,balanced,0.08753599723180135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,64,balanced,0.0976106623808543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,64,balanced,0.11702932914098103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,64,balanced,0.1646666626135508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,64,balanced,0.14015466968218485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,64,balanced,0.17805866400400797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,64,balanced,0.2076373298962911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,64,balanced,0.21022399266560873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,64,balanced,0.28230400880177814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,64,balanced,0.24546132485071817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,64,balanced,0.3505386511484782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,64,balanced,0.32899733384450275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,64,balanced,0.4917546510696411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,64,balanced,0.40770665804545086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,64,balanced,0.6295466820398966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,64,balanced,0.05230399966239929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,64,balanced,0.05343466500441233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,64,balanced,0.7746026515960693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,64,balanced,0.053264002005259194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,64,balanced,0.05682666599750519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,64,balanced,0.5057546695073446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,64,balanced,0.05884799857934316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,64,balanced,0.069733331600825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,64,balanced,0.0673280010620753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,64,balanced,0.06804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,64,balanced,0.06739733119805653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,64,balanced,0.06877333422501881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,64,balanced,0.06497600177923839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,64,balanced,0.06738133231798808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,64,balanced,0.07060266534487407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,64,balanced,0.06901866694291432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,64,balanced,0.07447466750939687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,64,balanced,0.07612800101439159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,64,balanced,0.02183466653029124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,64,balanced,0.07747200131416321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,64,balanced,0.024405332903067272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,64,balanced,0.08334400256474812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,64,balanced,0.0232640008131663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,64,balanced,0.023999998966852825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,64,balanced,0.090037335952123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,64,balanced,0.02443733314673106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,64,balanced,0.028416000306606293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,64,balanced,0.05349333087603251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,64,balanced,0.10205333431561787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,64,balanced,0.05346133311589559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,64,balanced,1.1873973210652669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,64,balanced,0.04446400205294291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,64,balanced,0.11288000146547954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,64,balanced,0.04574400186538696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,64,balanced,0.04530133306980133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,64,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,64,balanced,0.14478400349617004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,64,balanced,0.042170668641726174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,64,balanced,0.043621331453323364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,64,balanced,0.052570665876070656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,64,balanced,0.7500106493631998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,64,balanced,0.05152533451716105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,64,balanced,0.16679465770721436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,64,balanced,0.05144000053405762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,64,balanced,0.05685866872469584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,64,balanced,0.06694933275381725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,64,balanced,0.22022932767868042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,64,balanced,0.09036266803741455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,64,balanced,0.10229866703351338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,64,balanced,0.26920533180236816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,64,balanced,0.141157329082489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,64,balanced,0.17673067251841226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,64,balanced,0.2497226595878601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,64,balanced,0.37004268169403076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,64,balanced,0.31248533725738525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,64,balanced,0.47014931837717694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,64,balanced,2.3382293383280435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,64,balanced,0.4623946746190389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,64,balanced,0.5724960168202718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,64,balanced,0.6178986628850301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,64,balanced,1.455743948618571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,64,balanced,0.7690666516621908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,64,balanced,0.8699466387430826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,64,balanced,1.2128106753031414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,64,balanced,1.7008800506591797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,64,balanced,2.390613396962484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,64,power_law_1.01,0.0707647979259491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,64,power_law_1.01,0.06863359808921814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,64,power_law_1.01,0.06446719765663148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,64,power_law_1.01,0.07128319740295411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,64,power_law_1.01,0.06887680292129517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,64,power_law_1.01,0.07123200297355652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,64,power_law_1.01,0.07455360293388366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,64,power_law_1.01,0.07208319902420043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,64,power_law_1.01,0.05928320288658142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,64,power_law_1.01,0.07624319791793824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,64,power_law_1.01,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,64,power_law_1.01,0.07637760043144226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,64,power_law_1.01,0.055270397663116456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,64,power_law_1.01,0.07601280212402343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,64,power_law_1.01,0.05612800121307373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,64,power_law_1.01,0.07749760150909424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,64,power_law_1.01,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,64,power_law_1.01,0.07825919985771179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,64,power_law_1.01,0.06216959953308106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,64,power_law_1.01,0.07922559976577759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,64,power_law_1.01,0.0642624020576477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,64,power_law_1.01,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,64,power_law_1.01,0.06292480230331421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,64,power_law_1.01,0.08588160276412964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,64,power_law_1.01,0.06424319744110107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,64,power_law_1.01,0.0868224024772644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,64,power_law_1.01,0.06376960277557372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,64,power_law_1.01,0.09801599979400635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,64,power_law_1.01,0.06561279892921448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,64,power_law_1.01,0.0654528021812439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,64,power_law_1.01,0.10451200008392333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,64,power_law_1.01,0.0673471987247467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,64,power_law_1.01,0.12159359455108643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,64,power_law_1.01,0.06826879978179931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,64,power_law_1.01,0.13415039777755738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,64,power_law_1.01,0.07405440211296081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,64,power_law_1.01,0.16062079668045043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,64,power_law_1.01,0.07404800057411194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,64,power_law_1.01,0.20465919971466065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,64,power_law_1.01,0.07901440262794494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,64,power_law_1.01,0.2533951997756958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,64,power_law_1.01,0.08971520066261292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,64,power_law_1.01,0.3277184009552002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,64,power_law_1.01,0.10300159454345703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,64,power_law_1.01,0.4505727767944336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,64,power_law_1.01,0.12624640464782716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,64,power_law_1.01,0.5680895805358886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,64,power_law_1.01,0.14595839977264405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,64,power_law_1.01,0.7260735988616943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,64,power_law_1.01,0.18970240354537965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,64,power_law_1.01,1.0486528396606445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,64,power_law_1.01,0.2314687967300415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,64,power_law_1.01,0.3224128007888794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,64,power_law_1.01,2.2033407211303713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,64,power_law_1.01,0.4113408088684082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,64,power_law_1.01,0.5776319980621338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,64,power_law_1.01,0.7607423782348632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,64,power_law_1.01,0.9133631706237793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,64,power_law_1.01,1.4352319717407227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,64,power_law_1.01,2.8962879180908203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,64,power_law_1.01,0.063673597574234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,64,power_law_1.01,0.06125440001487732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,64,power_law_1.01,0.05719040036201477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,64,power_law_1.01,0.05618559718132019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,64,power_law_1.01,0.058719998598098753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,64,power_law_1.01,0.0636672019958496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,64,power_law_1.01,0.06373760104179382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,64,power_law_1.01,0.06766719818115234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,64,power_law_1.01,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,64,power_law_1.01,0.06632320284843445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,64,power_law_1.01,0.06685439944267273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,64,power_law_1.01,0.02107519954442978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,64,power_law_1.01,0.06986240148544312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,64,power_law_1.01,0.021324799954891206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,64,power_law_1.01,0.06587520241737366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,64,power_law_1.01,0.02237440049648285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,64,power_law_1.01,0.06836479902267456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,64,power_law_1.01,0.02131199985742569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,64,power_law_1.01,0.07395840287208558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,64,power_law_1.01,0.02232320010662079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,64,power_law_1.01,0.07284479737281799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,64,power_law_1.01,0.02629759907722473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,64,power_law_1.01,0.0763264000415802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,64,power_law_1.01,0.05181440114974976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,64,power_law_1.01,0.051660799980163576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,64,power_law_1.01,0.08624640107154846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,64,power_law_1.01,0.043609601259231565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.01,0.09285759925842285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,64,power_law_1.01,0.042828801274299624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.01,0.11255040168762206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,64,power_law_1.01,0.04360319972038269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.01,0.12947839498519897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,64,power_law_1.01,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.01,0.16587519645690918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,64,power_law_1.01,0.043910399079322815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.01,0.1878335952758789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,64,power_law_1.01,0.042771199345588685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.01,0.26105599403381347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,64,power_law_1.01,0.043705600500106814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.01,0.3250047922134399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,64,power_law_1.01,0.04417920112609863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.01,0.4701119899749756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,64,power_law_1.01,0.0471231997013092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.01,0.5932735919952392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,64,power_law_1.01,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,64,power_law_1.01,0.703334379196167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.01,0.05537279844284058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,64,power_law_1.01,1.1150143623352051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.01,0.06535040140151978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.01,0.07440639734268188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,64,power_law_1.01,2.271628761291504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.01,0.10602879524230957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.01,0.12821120023727417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.01,0.1739583969116211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.01,0.20873599052429198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.01,0.28517119884490966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.01,0.37204480171203613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,64,power_law_1.01,0.47694082260131837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,64,power_law_1.01,0.8313535690307617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,64,power_law_1.01,1.4845824241638184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,64,power_law_1.2,0.07288960218429566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,64,power_law_1.2,0.07100800275802613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,64,power_law_1.2,0.06422399878501892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,64,power_law_1.2,0.06937599778175355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,64,power_law_1.2,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,64,power_law_1.2,0.07272959947586059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,64,power_law_1.2,0.07560960054397584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,64,power_law_1.2,0.07494400143623352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,64,power_law_1.2,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,64,power_law_1.2,0.07759360074996949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,64,power_law_1.2,0.07784960269927979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,64,power_law_1.2,0.061459201574325564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,64,power_law_1.2,0.05813760161399841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,64,power_law_1.2,0.07661439776420594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,64,power_law_1.2,0.05479679703712463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,64,power_law_1.2,0.07587199807167053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,64,power_law_1.2,0.056966400146484374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,64,power_law_1.2,0.08002560138702393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,64,power_law_1.2,0.05877760052680969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,64,power_law_1.2,0.08605440258979798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,64,power_law_1.2,0.085971200466156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,64,power_law_1.2,0.06097279787063599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,64,power_law_1.2,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,64,power_law_1.2,0.09251199960708618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,64,power_law_1.2,0.06471040248870849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,64,power_law_1.2,0.09976959824562073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,64,power_law_1.2,0.0664255976676941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,64,power_law_1.2,0.10773119926452637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,64,power_law_1.2,0.06520959734916687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,64,power_law_1.2,0.12595839500427247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,64,power_law_1.2,0.06630399823188782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,64,power_law_1.2,0.13882240056991577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,64,power_law_1.2,0.06517120003700257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,64,power_law_1.2,0.17864320278167725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,64,power_law_1.2,0.06746240258216858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,64,power_law_1.2,0.2198335886001587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,64,power_law_1.2,0.06811519861221313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,64,power_law_1.2,0.28333439826965334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,64,power_law_1.2,0.07175679802894593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,64,power_law_1.2,0.3771712064743042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,64,power_law_1.2,0.07431679964065552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,64,power_law_1.2,0.5564415931701661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,64,power_law_1.2,0.08184959888458251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,64,power_law_1.2,0.7178175926208497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,64,power_law_1.2,0.09557120203971863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,64,power_law_1.2,0.8640128135681152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,64,power_law_1.2,0.1055616021156311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,64,power_law_1.2,1.4541184425354003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,64,power_law_1.2,0.13295359611511232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,64,power_law_1.2,0.15366400480270387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,64,power_law_1.2,2.737721633911133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,64,power_law_1.2,0.1987264037132263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,64,power_law_1.2,0.24983038902282714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,64,power_law_1.2,0.3395456075668335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,64,power_law_1.2,0.45424637794494627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,64,power_law_1.2,0.6810111999511719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,64,power_law_1.2,0.8959936141967774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,64,power_law_1.2,1.140281581878662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,64,power_law_1.2,1.7379840850830077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,64,power_law_1.2,3.587839889526367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,64,power_law_1.2,0.06273279786109924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,64,power_law_1.2,0.06476799845695495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,64,power_law_1.2,0.053388798236846925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,64,power_law_1.2,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,64,power_law_1.2,0.05886719822883606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,64,power_law_1.2,0.020979200303554536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,64,power_law_1.2,0.06266239881515503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,64,power_law_1.2,0.02197760045528412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,64,power_law_1.2,0.06373760104179382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,64,power_law_1.2,0.06542720198631287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,64,power_law_1.2,0.02213120013475418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,64,power_law_1.2,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,64,power_law_1.2,0.021631999313831328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,64,power_law_1.2,0.06418560147285461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,64,power_law_1.2,0.02295680046081543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,64,power_law_1.2,0.06632959842681885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,64,power_law_1.2,0.027084800601005554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,64,power_law_1.2,0.067084801197052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,64,power_law_1.2,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,64,power_law_1.2,0.06792320013046264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,64,power_law_1.2,0.05199360251426697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,64,power_law_1.2,0.06826879978179931
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,64,power_law_1.2,0.04176000058650971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,64,power_law_1.2,0.043347200751304625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,64,power_law_1.2,0.07479680180549622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,64,power_law_1.2,0.07372159957885742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,64,power_law_1.2,0.04281600117683411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,64,power_law_1.2,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,64,power_law_1.2,0.07852799892425537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,64,power_law_1.2,0.04292480051517487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,64,power_law_1.2,0.08696960210800171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,64,power_law_1.2,0.04322560131549835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.2,0.09498239755630493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,64,power_law_1.2,0.04322560131549835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.2,0.1159551978111267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,64,power_law_1.2,0.04462080001831055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.2,0.13151999711990356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,64,power_law_1.2,0.04711039960384369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.2,0.17022080421447755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,64,power_law_1.2,0.05059199929237366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.2,0.20475521087646484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.2,0.2927232027053833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.2,0.06718720197677612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.2,0.36494081020355223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.2,0.07493759989738465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.2,0.5726975917816162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.2,0.10697599649429321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.2,0.691648006439209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.2,0.12603520154953002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,64,power_law_1.2,0.85098876953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.2,0.18224639892578126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,64,power_law_1.2,1.3362815856933594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.2,0.2371583938598633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,64,power_law_1.2,2.846086311340332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.2,0.31962881088256834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.2,0.4439807891845703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,64,power_law_1.2,0.5473408222198486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,64,power_law_1.2,0.8372096061706543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,64,power_law_1.2,1.7565824508666992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,32,balanced,0.05713599920272827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,32,balanced,0.058058664202690125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,32,balanced,0.05823466678460439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,32,balanced,0.061946665247281395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,32,balanced,0.06579199930032094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,32,balanced,0.07644799848397572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,32,balanced,0.07573866844177246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,32,balanced,0.07801066835721333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,32,balanced,0.07989866534868877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,32,balanced,0.08001600205898285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,32,balanced,0.07748266557852428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,32,balanced,0.07869333525498708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,32,balanced,0.08102400104204814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,32,balanced,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,32,balanced,0.05161066850026449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,32,balanced,0.07787733276685078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,32,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,32,balanced,0.09028266867001851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,32,balanced,0.057392001152038574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,32,balanced,0.057392001152038574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,32,balanced,0.08754133184750874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,32,balanced,0.06544533371925354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,32,balanced,0.09099733829498291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,32,balanced,0.06577600042025249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,32,balanced,0.06755733489990234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,32,balanced,0.09941866993904114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,32,balanced,0.06807999809583028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,32,balanced,0.06967466572920482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,32,balanced,0.1027946670850118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,32,balanced,0.06810133159160614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,32,balanced,0.06754133105278015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,32,balanced,0.11674132943153381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,32,balanced,0.0710399995247523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,32,balanced,0.07133333384990692
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,32,balanced,0.075573335091273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,32,balanced,0.12719466288884482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,32,balanced,0.07727466523647308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,32,balanced,0.08120533327261607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,32,balanced,0.15153066317240396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,32,balanced,0.08941866954167683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,32,balanced,0.0999893347422282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,32,balanced,0.12366933623949687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,32,balanced,0.17150932550430298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,32,balanced,0.14525333046913147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,32,balanced,0.18337066968282065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,32,balanced,0.21707199017206827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,32,balanced,0.21825067202250162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,32,balanced,0.2959360082944234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,32,balanced,0.2595199942588806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,32,balanced,0.36481066544850665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,32,balanced,0.3480000098546346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,32,balanced,0.052522664268811546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,32,balanced,0.05329066514968872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,32,balanced,0.05327466626962026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,32,balanced,0.5134079853693644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,32,balanced,0.0609493354956309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,32,balanced,0.061530664563179016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,32,balanced,0.06734399994214375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,32,balanced,0.06771733363469441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,32,balanced,0.06740800042947133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,32,balanced,0.02146666745344798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,32,balanced,0.06919999917348225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,32,balanced,0.023930666347344715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,32,balanced,0.06931200126806895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,32,balanced,0.02402133246262868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,32,balanced,0.07038400073846181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,32,balanced,0.02404800057411194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,32,balanced,0.023957334458827972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,32,balanced,0.0710453341404597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,32,balanced,0.024271999796231587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,32,balanced,0.07177599767843883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,32,balanced,0.43295466899871826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,32,balanced,0.030576000610987347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,32,balanced,0.6575626532236735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,32,balanced,0.07181333502133687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,32,balanced,0.02861333390076955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,32,balanced,0.07884799937407176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,32,balanced,0.05533333122730255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,32,balanced,0.07887466748555501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,32,balanced,0.05491200089454651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,32,balanced,0.05526933570702871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,32,balanced,0.08316799998283386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,32,balanced,0.048581331968307495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,32,balanced,0.08754666646321614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,32,balanced,0.046767999728520714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,32,balanced,0.04655466477076212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,32,balanced,0.09492266178131104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,32,balanced,0.054976001381874084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,32,balanced,0.10641599694887798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,32,balanced,0.054832001527150474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,32,balanced,0.05293866495291392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,32,balanced,0.12227200468381245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,32,balanced,0.06136000156402588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,32,balanced,0.8084906737009684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,32,balanced,0.0689386675755183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,32,balanced,0.1521013379096985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,32,balanced,0.5406560103098551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,32,balanced,0.0792746643225352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,32,balanced,0.0922826627890269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,32,balanced,0.1775253415107727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,32,balanced,0.1304213305314382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,32,balanced,0.23397332429885864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,32,balanced,0.15599466363588968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,32,balanced,0.2850826581319173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,32,balanced,0.21416000525156656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,32,balanced,1.2427466710408528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,32,balanced,0.2723413308461507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,32,balanced,0.39501333236694336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,32,balanced,0.7978400389353434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,32,balanced,0.39795200030008954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,32,balanced,0.50163201491038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,32,balanced,0.5210293531417847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,32,balanced,0.6083306471506754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,32,balanced,0.6464159886042277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,32,balanced,2.4556585947672525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,32,balanced,0.9280746777852377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,32,balanced,1.0297973155975342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,32,balanced,1.5608213742574055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,32,balanced,1.831925392150879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,32,balanced,2.008672078450521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,32,power_law_1.01,0.0779263973236084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,32,power_law_1.01,0.07477759718894958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,32,power_law_1.01,0.06772480010986329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,32,power_law_1.01,0.0673471987247467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,32,power_law_1.01,0.06588159799575806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,32,power_law_1.01,0.06782720088958741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,32,power_law_1.01,0.06258559823036194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,32,power_law_1.01,0.07534080147743225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,32,power_law_1.01,0.05576320290565491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,32,power_law_1.01,0.0748799979686737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,32,power_law_1.01,0.05731199979782105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,32,power_law_1.01,0.07919359803199769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,32,power_law_1.01,0.0602944016456604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,32,power_law_1.01,0.061689597368240354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,32,power_law_1.01,0.07978240251541138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,32,power_law_1.01,0.06471040248870849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,32,power_law_1.01,0.08077440261840821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,32,power_law_1.01,0.06455039978027344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,32,power_law_1.01,0.08030080199241638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,32,power_law_1.01,0.06554880142211914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,32,power_law_1.01,0.08048639893531799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,32,power_law_1.01,0.06499199867248535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,32,power_law_1.01,0.08117120265960694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,32,power_law_1.01,0.0659712016582489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,32,power_law_1.01,0.08350719809532166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,32,power_law_1.01,0.06835839748382569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,32,power_law_1.01,0.08903040289878845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,32,power_law_1.01,0.06890239715576171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,32,power_law_1.01,0.08956800103187561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,32,power_law_1.01,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,32,power_law_1.01,0.09293439984321594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,32,power_law_1.01,0.07555199861526489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,32,power_law_1.01,0.1028480052947998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,32,power_law_1.01,0.07485439777374267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,32,power_law_1.01,0.1095039963722229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,32,power_law_1.01,0.08283519744873047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,32,power_law_1.01,0.12462719678878784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,32,power_law_1.01,0.09432960152626038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,32,power_law_1.01,0.13699840307235717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,32,power_law_1.01,0.10808320045471191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,32,power_law_1.01,0.17129600048065186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,32,power_law_1.01,0.1352959990501404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,32,power_law_1.01,0.1938815951347351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,32,power_law_1.01,0.1571328043937683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,32,power_law_1.01,0.2587968111038208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,32,power_law_1.01,0.20350079536437987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,32,power_law_1.01,0.3301440000534058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,32,power_law_1.01,0.23900160789489747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,32,power_law_1.01,0.46248321533203124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,32,power_law_1.01,0.33844480514526365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,32,power_law_1.01,0.5930496215820312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,32,power_law_1.01,0.4213247776031494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,32,power_law_1.01,0.6854400157928466
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,32,power_law_1.01,0.021561600267887115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,32,power_law_1.01,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,32,power_law_1.01,0.5949120044708252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,32,power_law_1.01,1.0828288078308106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,32,power_law_1.01,0.02170239984989166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,32,power_law_1.01,0.06696320176124573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,32,power_law_1.01,0.7804416179656982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,32,power_law_1.01,0.021792000532150267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,32,power_law_1.01,0.060121601819992064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,32,power_law_1.01,2.166598320007324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,32,power_law_1.01,0.9505472183227539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,32,power_law_1.01,0.022438399493694305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,32,power_law_1.01,0.06032639741897583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,32,power_law_1.01,0.02184319943189621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,32,power_law_1.01,0.060096001625061034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,32,power_law_1.01,1.4871935844421387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,32,power_law_1.01,0.022316800057888032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,32,power_law_1.01,0.06360960006713867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,32,power_law_1.01,2.9557695388793945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,32,power_law_1.01,0.027718400955200194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,32,power_law_1.01,0.066457599401474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,32,power_law_1.01,0.06888319849967957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,32,power_law_1.01,0.028339201211929323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,32,power_law_1.01,0.05233280062675476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,32,power_law_1.01,0.06825600266456604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,32,power_law_1.01,0.05235840082168579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,32,power_law_1.01,0.07010560035705567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,32,power_law_1.01,0.053324800729751584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,32,power_law_1.01,0.06819199919700622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,32,power_law_1.01,0.044940799474716187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,32,power_law_1.01,0.06981120109558106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,32,power_law_1.01,0.04442239999771118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,32,power_law_1.01,0.07098879814147949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,32,power_law_1.01,0.044409599900245664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,32,power_law_1.01,0.07368959784507752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,32,power_law_1.01,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,32,power_law_1.01,0.07767040133476258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,32,power_law_1.01,0.0777728021144867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,32,power_law_1.01,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,32,power_law_1.01,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,32,power_law_1.01,0.08193920254707336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,32,power_law_1.01,0.05310720205307007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,32,power_law_1.01,0.08835840225219727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.01,0.09941759705543518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.01,0.05813760161399841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.01,0.06758400201797485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.01,0.11816320419311524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.01,0.07690879702568054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.01,0.13640960454940795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.01,0.09664639830589294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.01,0.1719231963157654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.01,0.11478400230407715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.01,0.20597119331359864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.01,0.16372480392456054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.01,0.2819135904312134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.01,0.19701759815216063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.01,0.34114561080932615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.01,0.2824640035629272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.01,0.4888576030731201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.01,0.6225088119506836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.01,0.36029438972473143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,32,power_law_1.01,0.4587071895599365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,32,power_law_1.01,0.7493504047393799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,32,power_law_1.01,1.2437503814697266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,32,power_law_1.01,0.6942527770996094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,32,power_law_1.01,2.4572416305541993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,32,power_law_1.01,1.3008447647094727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,32,power_law_1.2,0.06506879925727845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,32,power_law_1.2,0.07927680015563965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,32,power_law_1.2,0.06093440055847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,32,power_law_1.2,0.07585279941558838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,32,power_law_1.2,0.05134720206260681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,32,power_law_1.2,0.06447359919548035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,32,power_law_1.2,0.0559935986995697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,32,power_law_1.2,0.07215359807014465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,32,power_law_1.2,0.05806080102920532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,32,power_law_1.2,0.07148799896240235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,32,power_law_1.2,0.06069759726524353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,32,power_law_1.2,0.07523199915885925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,32,power_law_1.2,0.06399359703063964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,32,power_law_1.2,0.07825919985771179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,32,power_law_1.2,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,32,power_law_1.2,0.07878400087356567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,32,power_law_1.2,0.06483839750289917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,32,power_law_1.2,0.07886080145835876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,32,power_law_1.2,0.06584320068359376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,32,power_law_1.2,0.07971199750900268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,32,power_law_1.2,0.0670144021511078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,32,power_law_1.2,0.08044800162315369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,32,power_law_1.2,0.06817920207977295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,32,power_law_1.2,0.08091520071029663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,32,power_law_1.2,0.07029119729995728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,32,power_law_1.2,0.08209279775619507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,32,power_law_1.2,0.06986879706382751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,32,power_law_1.2,0.08257279992103576
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,32,power_law_1.2,0.07432320117950439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,32,power_law_1.2,0.08807039856910706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,32,power_law_1.2,0.07560319900512695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,32,power_law_1.2,0.08945919871330262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,32,power_law_1.2,0.08617600202560424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,32,power_law_1.2,0.09513599872589111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,32,power_law_1.2,0.11084799766540528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,32,power_law_1.2,0.13864959478378297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,32,power_law_1.2,0.16218240261077882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,32,power_law_1.2,0.20648961067199706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,32,power_law_1.2,0.25134079456329345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,32,power_law_1.2,0.07017599940299987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,32,power_law_1.2,0.3760895967483521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,32,power_law_1.2,0.0667136013507843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,32,power_law_1.2,0.05539839863777161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,32,power_law_1.2,0.4763008117675781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,32,power_law_1.2,0.06037120223045349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,32,power_law_1.2,0.021823999285697938
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,32,power_law_1.2,0.6461696147918701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,32,power_law_1.2,0.05994240045547485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,32,power_law_1.2,0.022643199563026427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,32,power_law_1.2,0.8581631660461426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,32,power_law_1.2,0.06244480013847351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,32,power_law_1.2,0.021958400309085847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,32,power_law_1.2,1.0878080368041991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,32,power_law_1.2,0.0686016023159027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,32,power_law_1.2,0.023001599311828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,32,power_law_1.2,1.6908863067626954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,32,power_law_1.2,0.06548479795455933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,32,power_law_1.2,0.02210559993982315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,32,power_law_1.2,0.06781439781188965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,32,power_law_1.2,0.022431999444961548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,32,power_law_1.2,3.355846405029297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,32,power_law_1.2,0.06898559927940369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,32,power_law_1.2,0.028883200883865357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,32,power_law_1.2,0.07052159905433655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,32,power_law_1.2,0.028147199749946596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,32,power_law_1.2,0.07440000176429748
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,32,power_law_1.2,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,32,power_law_1.2,0.07160320281982421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,32,power_law_1.2,0.053439998626708986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,32,power_law_1.2,0.07466880083084107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,32,power_law_1.2,0.05403519868850708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,32,power_law_1.2,0.0766207993030548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,32,power_law_1.2,0.04593920111656189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,32,power_law_1.2,0.07992960214614868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,32,power_law_1.2,0.045440000295639035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,32,power_law_1.2,0.08239359855651855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,32,power_law_1.2,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,32,power_law_1.2,0.09364479780197144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,32,power_law_1.2,0.04813440144062042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.2,0.10296319723129273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,32,power_law_1.2,0.04737280011177063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.2,0.11914880275726318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,32,power_law_1.2,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.2,0.1431872010231018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,32,power_law_1.2,0.05271040201187134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.2,0.17870080471038818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.2,0.06048640012741089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.2,0.21804800033569335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.2,0.07052159905433655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.2,0.29453439712524415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.2,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.2,0.39562880992889404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.2,0.10042879581451417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.2,0.5533376216888428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.2,0.12265599966049194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.2,0.7279679775238037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.2,0.1664255976676941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,32,power_law_1.2,0.905459213256836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.2,0.20821120738983154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.2,0.2980736017227173
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,32,power_law_1.2,1.4076416015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.2,0.3765887975692749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,32,power_law_1.2,2.769222450256348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,32,power_law_1.2,0.4933568000793457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,32,power_law_1.2,0.7317503929138184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,32,power_law_1.2,1.5249407768249512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,32,power_law_1.2,0.09303680062294006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,16,balanced,0.054485330979029335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,16,balanced,0.05710400144259135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,32,power_law_1.2,0.10540159940719604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,16,balanced,0.05880000193913778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,32,power_law_1.2,0.11297919750213622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,16,balanced,0.06509333352247874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,16,balanced,0.06756799916426341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,32,power_law_1.2,0.12922240495681764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,16,balanced,0.08002666632334392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,16,balanced,0.08155733346939087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,32,power_law_1.2,0.1441472053527832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,16,balanced,0.08050666749477386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,32,power_law_1.2,0.17454080581665038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,16,balanced,0.08398933211962382
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,32,power_law_1.2,0.2080832004547119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,16,balanced,0.08624000350634257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,16,balanced,0.08014399806658427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,32,power_law_1.2,0.2867520093917847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,16,balanced,0.08618133266766866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,32,power_law_1.2,0.35948801040649414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,16,balanced,0.08552533388137817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,32,power_law_1.2,0.49881601333618164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,16,balanced,0.0876533289750417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,16,balanced,0.09202667077382405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,32,power_law_1.2,0.7234879970550537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,16,balanced,0.0960586667060852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,32,power_law_1.2,0.9192768096923828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,16,balanced,0.09805867075920105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,32,power_law_1.2,1.3061375617980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,16,balanced,0.10446400443712871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,32,power_law_1.2,2.7683263778686524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,16,balanced,0.11132799585660298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,16,balanced,0.12544000148773193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,16,balanced,0.050341332952181496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,16,balanced,0.13662399848302206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,16,balanced,0.052442664901415505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,16,balanced,0.053157334526379905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,16,balanced,0.05780800183614095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,16,balanced,0.06333333253860474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,16,balanced,0.16219199697176614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,16,balanced,0.06950399776299794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,16,balanced,0.06942399839560191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,16,balanced,0.0712960014740626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,16,balanced,0.07165333131949107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,16,balanced,0.18393067518870035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,16,balanced,0.052330667773882546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,16,balanced,0.0711893339951833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,16,balanced,0.05338666836420695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,16,balanced,0.07208533088366191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,16,balanced,0.05718400080998739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,16,balanced,0.07293333113193512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,16,balanced,0.05889600018660227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,16,balanced,0.07340266803900401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,16,balanced,0.06804266571998596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,16,balanced,0.07607999940713246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,16,balanced,0.24172266324361166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,16,balanced,0.07653333246707916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,16,balanced,0.08003200093905131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,16,balanced,0.07941866914431255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,16,balanced,0.08317333459854126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,16,balanced,0.08146133522192638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,16,balanced,0.07928533355395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,16,balanced,0.08874666690826416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,16,balanced,0.07949333389600118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,16,balanced,0.07937066753705342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,16,balanced,0.09713066617647807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,16,balanced,0.2876266638437907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,16,balanced,0.08165333171685536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,16,balanced,0.10777599612871806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,16,balanced,0.08146666487058003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,16,balanced,0.08585066596666972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,16,balanced,0.13361600041389465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,16,balanced,0.08810133735338847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,16,balanced,0.08944533268610637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,16,balanced,0.1565546691417694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,16,balanced,0.09302399555842082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,16,balanced,0.10268266995747884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,16,balanced,0.39267198244730633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,16,balanced,0.19631999731063843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,16,balanced,0.10633599758148193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,16,balanced,0.12493866682052612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,16,balanced,0.23329599698384604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,16,balanced,0.14111999670664468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,16,balanced,0.17252800861994425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,16,balanced,0.3184106747309367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,16,balanced,0.4938240051269531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,16,balanced,0.20014933745066324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,16,balanced,0.3930186827977498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,16,balanced,0.26764265696207684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,16,balanced,0.6133600076039633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,16,balanced,0.556442658106486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,16,balanced,0.32264532645543414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,16,balanced,0.7129813035329183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,16,balanced,0.45366934935251874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,16,balanced,0.9174986680348715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,16,balanced,0.5751680135726929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,16,balanced,0.8841599623362223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,16,balanced,0.02370133250951767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,16,balanced,0.708021322886149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,16,balanced,0.02441066751877467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,16,balanced,0.024864000578721363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,16,balanced,0.0242399995525678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,16,balanced,0.027722666660944622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,16,balanced,0.026549334327379864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,16,balanced,0.0283146674434344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,16,balanced,0.02834133307139079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,16,balanced,1.3595892588297527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,16,balanced,0.035877334574858345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,16,balanced,0.036831999818483986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,16,balanced,0.034517332911491394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,16,balanced,0.0595413347085317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,16,balanced,0.059343998630841575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,16,balanced,0.060090666015942894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,16,balanced,0.0566293348868688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,16,balanced,0.05866133173306783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,16,balanced,0.059338668982187905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,16,balanced,0.06703466673692067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,16,balanced,0.07470400134722392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,16,balanced,0.08813333511352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,16,balanced,1.0912106831868489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,16,balanced,0.10161067048708598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,16,balanced,1.7872692743937175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,16,balanced,0.12546666463216147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,16,balanced,0.15887999534606934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,16,balanced,0.21034133434295654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,16,balanced,0.2559199929237366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,16,balanced,2.6809921264648438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,16,balanced,0.3743520180384318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,16,balanced,0.4830506642659505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,16,balanced,2.133530616760254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,16,balanced,0.605626662572225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,16,balanced,0.957365353902181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,16,power_law_1.01,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,16,power_law_1.01,0.08377599716186523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,16,power_law_1.01,0.066457599401474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,16,power_law_1.01,0.0695039987564087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,16,power_law_1.01,0.07247359752655029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,16,power_law_1.01,0.07797120213508606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,16,power_law_1.01,0.08132479786872863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,16,power_law_1.01,0.08353919982910156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,16,power_law_1.01,0.08399999737739564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,16,power_law_1.01,0.08240640163421631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,16,power_law_1.01,0.08409600257873535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,16,power_law_1.01,0.08638719916343689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,16,power_law_1.01,0.08727040290832519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,16,power_law_1.01,0.08752639889717102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,16,power_law_1.01,0.09288960099220275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,16,power_law_1.01,0.09333119988441467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,16,power_law_1.01,0.09836800098419189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,16,power_law_1.01,0.10870399475097656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,16,balanced,1.8944692611694336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,16,power_law_1.01,0.11564160585403442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,16,power_law_1.01,0.06563839912414551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,16,power_law_1.01,0.13608319759368898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,16,power_law_1.01,0.07384960055351257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,16,power_law_1.01,0.15005439519882202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,16,power_law_1.01,0.055731201171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,16,power_law_1.01,0.18316160440444945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,16,power_law_1.01,0.05848960280418396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,16,power_law_1.01,0.21647360324859619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,16,power_law_1.01,0.27850239276885985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,16,power_law_1.01,0.060915201902389526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,16,power_law_1.01,0.3409087896347046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,16,power_law_1.01,0.0646336019039154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,16,power_law_1.01,0.0677183985710144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,16,power_law_1.01,0.5088255882263184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,16,power_law_1.01,0.0700160026550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,16,power_law_1.01,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,16,power_law_1.01,0.64552321434021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,16,power_law_1.01,0.08241279721260071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,16,power_law_1.01,0.0686079978942871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,16,power_law_1.01,0.05802239775657654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,16,power_law_1.01,0.8278592109680176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,16,power_law_1.01,0.07043200135231018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,16,power_law_1.01,0.062105602025985716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,16,power_law_1.01,1.2770943641662598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,16,power_law_1.01,0.06476799845695495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,16,power_law_1.01,0.07212799787521362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,16,power_law_1.01,0.06752640008926392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,16,power_law_1.01,0.07178879976272583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,16,power_law_1.01,2.4992128372192384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,16,power_law_1.01,0.07731840014457703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,16,power_law_1.01,0.07297279834747314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,16,power_law_1.01,0.07406079769134521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,16,power_law_1.01,0.0770367980003357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,16,power_law_1.01,0.07694720029830933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,16,power_law_1.01,0.08109440207481385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,16,power_law_1.01,0.07875199913978577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,16,power_law_1.01,0.0843775987625122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,16,power_law_1.01,0.07779840230941773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,16,power_law_1.01,0.09132800102233887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,16,power_law_1.01,0.07987200021743775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,16,power_law_1.01,0.10592639446258545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,16,power_law_1.01,0.07934079766273498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,16,power_law_1.01,0.11798399686813354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,16,power_law_1.01,0.08252800107002259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,16,power_law_1.01,0.14338560104370118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,16,power_law_1.01,0.08293759822845459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,16,power_law_1.01,0.16974079608917236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,16,power_law_1.01,0.08791040182113648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,16,power_law_1.01,0.21760001182556152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,16,power_law_1.01,0.09247360229492188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,16,power_law_1.01,0.25603840351104734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,16,power_law_1.01,0.09986559748649597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,16,power_law_1.01,0.3610304117202759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.01,0.11431039571762085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,16,power_law_1.01,0.4391359806060791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.01,0.1383296012878418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,16,power_law_1.01,0.6539775848388671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.01,0.15225600004196166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,16,power_law_1.01,0.8558976173400878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.01,0.19605120420455932
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,16,power_law_1.01,1.016057586669922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.01,0.23936638832092286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.01,0.3110527992248535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,16,power_law_1.01,1.6323904037475585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.01,0.39680640697479247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,16,power_law_1.01,3.4458686828613283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.01,0.5568319797515869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.01,0.7202176094055176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,16,power_law_1.01,0.8783935546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,16,power_law_1.01,1.4342080116271974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,16,power_law_1.01,2.8310720443725588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,16,power_law_1.01,0.021376000344753267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,16,power_law_1.01,0.023001599311828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,16,power_law_1.01,0.022233599424362184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,16,power_law_1.01,0.02333440035581589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,16,power_law_1.01,0.02420479953289032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,16,power_law_1.01,0.023769600689411162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,16,power_law_1.01,0.025139200687408447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,16,power_law_1.01,0.025145599246025087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,16,power_law_1.01,0.03038719892501831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,16,power_law_1.01,0.030899199843406677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,16,power_law_1.01,0.03131519854068756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,16,power_law_1.01,0.055430400371551516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,16,power_law_1.01,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,16,power_law_1.01,0.05667200088500977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,16,power_law_1.01,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,16,power_law_1.01,0.052857601642608644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,16,power_law_1.01,0.054028797149658206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,16,power_law_1.01,0.05666559934616089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,16,power_law_1.2,0.07233920097351074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.01,0.06410880088806152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.01,0.07448959946632386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.01,0.0837823987007141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.01,0.1019327998161316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.01,0.11853439807891845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.01,0.1562559962272644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.01,0.20359039306640625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.01,0.2911808013916016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.01,0.3631808042526245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,16,power_law_1.01,0.45207037925720217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,16,power_law_1.01,0.6923711776733399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,16,power_law_1.01,1.3335871696472168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,16,power_law_1.2,0.06510080099105835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,16,power_law_1.2,0.06963840126991272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,16,power_law_1.2,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,16,power_law_1.2,0.06845440268516541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,16,power_law_1.2,0.05767040252685547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,16,power_law_1.2,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,16,power_law_1.2,0.06177279949188232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,16,power_law_1.2,0.057004797458648684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,16,power_law_1.2,0.06199679970741272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,16,power_law_1.2,0.06267520189285278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,16,power_law_1.2,0.06760960221290588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,16,power_law_1.2,0.0651968002319336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,16,power_law_1.2,0.06850559711456299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,16,power_law_1.2,0.06995840072631836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,16,power_law_1.2,0.06940159797668458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,16,power_law_1.2,0.0759552001953125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,16,power_law_1.2,0.06903679966926575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,16,power_law_1.2,0.07614079713821412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,16,power_law_1.2,0.07123200297355652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,16,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,16,power_law_1.2,0.07285760045051574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,16,power_law_1.2,0.07589120268821717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,16,power_law_1.2,0.0739135980606079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,16,power_law_1.2,0.07904639840126038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,16,power_law_1.2,0.07463039755821228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,16,power_law_1.2,0.07944959998130799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,16,power_law_1.2,0.08123520016670227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,16,power_law_1.2,0.07986559867858886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,16,power_law_1.2,0.08379520177841186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,16,power_law_1.2,0.08367999792098998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,16,power_law_1.2,0.08988159894943237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,16,power_law_1.2,0.08761600255966187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,16,power_law_1.2,0.10481280088424683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,16,power_law_1.2,0.08999040126800537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,16,power_law_1.2,0.12126719951629639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,16,power_law_1.2,0.09326080083847046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,16,power_law_1.2,0.14862719774246216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,16,power_law_1.2,0.10680960416793824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,16,power_law_1.2,0.17356799840927123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.2,0.11961599588394164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,16,power_law_1.2,0.22055680751800538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.2,0.13905919790267945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,16,power_law_1.2,0.266592001914978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.2,0.16071679592132568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,16,power_law_1.2,0.3763904094696045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.2,0.1982591986656189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,16,power_law_1.2,0.48798718452453616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.2,0.25594239234924315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.2,0.34253439903259275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,16,power_law_1.2,0.7133120059967041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.2,0.435916805267334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,16,power_law_1.2,0.9681920051574707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.2,0.6260287761688232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,16,power_law_1.2,1.190028762817383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.2,0.7920256137847901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,16,power_law_1.2,1.9208383560180664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,16,power_law_1.2,1.0689727783203125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,16,power_law_1.2,3.9086593627929687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,16,power_law_1.2,1.6052864074707032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,16,power_law_1.2,3.1727872848510743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,16,power_law_1.2,0.0787775993347168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,16,power_law_1.2,0.021190400421619415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,16,power_law_1.2,0.060063999891281125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,16,power_law_1.2,0.02245119959115982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,16,power_law_1.2,0.06581760048866273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,16,power_law_1.2,0.022495999932289124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,16,power_law_1.2,0.07102720141410827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,16,power_law_1.2,0.02327679991722107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,16,power_law_1.2,0.07605119943618774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,16,power_law_1.2,0.02470400035381317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,16,power_law_1.2,0.08023040294647217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,16,power_law_1.2,0.023948800563812257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,16,power_law_1.2,0.08236799836158752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,16,power_law_1.2,0.02459519952535629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,16,power_law_1.2,0.0840448021888733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,16,power_law_1.2,0.02550399899482727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,16,power_law_1.2,0.0846783995628357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,16,power_law_1.2,0.029260799288749695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,16,power_law_1.2,0.08279680013656616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,16,power_law_1.2,0.03022719919681549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,16,power_law_1.2,0.08638719916343689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,16,power_law_1.2,0.030854400992393494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,16,power_law_1.2,0.08762239813804626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,16,power_law_1.2,0.056966400146484374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,16,power_law_1.2,0.0884607970714569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,16,power_law_1.2,0.05559039711952209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,16,power_law_1.2,0.09470080137252808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,16,power_law_1.2,0.05639680027961731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,16,power_law_1.2,0.09523839950561523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,16,power_law_1.2,0.05234559774398804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,16,power_law_1.2,0.09960319995880126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,16,power_law_1.2,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,16,power_law_1.2,0.1103551983833313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,16,power_law_1.2,0.05303040146827698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,16,power_law_1.2,0.12075519561767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,16,power_law_1.2,0.05622400045394897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,16,power_law_1.2,0.13880319595336915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.2,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,16,power_law_1.2,0.15211520195007325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.2,0.0732096016407013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,16,power_law_1.2,0.18968960046768188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.2,0.0830784022808075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,16,power_law_1.2,0.22306559085845948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.2,0.10383360385894776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,16,power_law_1.2,0.31244161128997805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.2,0.12198400497436523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,16,power_law_1.2,0.3972991943359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.2,0.1712448000907898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,16,power_law_1.2,0.541484785079956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.2,0.21836159229278565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,16,power_law_1.2,0.7390272140502929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,8,balanced,0.05611200133959452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.2,0.28910720348358154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,8,balanced,0.05753066639105479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,16,power_law_1.2,0.9224896430969238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,8,balanced,0.0614026685555776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.2,0.38471040725708006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,16,power_law_1.2,1.4391231536865234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,8,balanced,0.0637600024541219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,8,balanced,0.0747626672188441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,16,power_law_1.2,0.47730560302734376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,8,balanced,0.09238933523495992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,16,power_law_1.2,2.992223930358887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,16,power_law_1.2,0.7480703830718994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,8,balanced,0.09055466453234355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,8,balanced,0.09319999814033508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,16,power_law_1.2,1.5291839599609376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,8,balanced,0.09363200267155965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,8,balanced,0.09407466650009155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,8,balanced,0.05018133421738943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,8,balanced,0.0516533354918162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,8,balanced,0.09231999516487122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,8,balanced,0.055402666330337524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,8,balanced,0.09582400321960449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,8,balanced,0.06163200239340464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,8,balanced,0.09649599591890971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,8,balanced,0.06807466844717662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,8,balanced,0.09872532884279887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,8,balanced,0.08155733346939087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,8,balanced,0.10564266641934712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,8,balanced,0.08265066643555959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,8,balanced,0.08226133386294048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,8,balanced,0.10479999581972758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,8,balanced,0.08357866605122884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,8,balanced,0.10979732871055603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,8,balanced,0.08387200037638347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,8,balanced,0.12010666728019714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,8,balanced,0.08168533444404602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,8,balanced,0.08621866504351298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,8,balanced,0.1251146694024404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,8,balanced,0.08713066577911377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,8,balanced,0.08620267113049825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,8,balanced,0.14331199725468954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,8,balanced,0.09344533085823059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,8,balanced,0.09532800316810608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,8,balanced,0.15596266587575278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,8,balanced,0.09835732976595561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,8,balanced,0.11417067050933838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,8,balanced,0.18552533785502115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,8,balanced,0.13032533725102743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,8,balanced,0.1561973293622335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,8,balanced,0.2148053248723348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,8,balanced,0.17619733015696207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,8,balanced,0.29014400641123456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,8,balanced,0.2214720050493876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,8,balanced,0.26401599248250324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,8,balanced,0.347327987353007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,8,balanced,0.3692266543706258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,8,balanced,0.484005331993103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,8,balanced,0.4596960147221883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,8,balanced,0.6154186725616455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,8,balanced,0.65611199537913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,8,balanced,0.7680213451385498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,8,balanced,0.8459626833597819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,8,balanced,1.0451040267944336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,8,balanced,1.156933307647705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,8,power_law_1.01,0.0651199996471405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,8,power_law_1.01,0.07303680181503296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,8,power_law_1.01,0.07058560252189636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,8,power_law_1.01,0.05919359922409058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,8,power_law_1.01,0.07799680233001709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,8,power_law_1.01,0.06281599998474122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,8,power_law_1.01,0.061868798732757566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,8,power_law_1.01,0.06672639846801758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,8,power_law_1.01,0.0692351996898651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,8,power_law_1.01,0.07249919772148132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,8,power_law_1.01,0.07489280104637146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,8,power_law_1.01,0.08249599933624267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,8,power_law_1.01,0.07737600207328796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,8,power_law_1.01,0.08496639728546143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,8,power_law_1.01,0.07799040079116822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,8,power_law_1.01,0.08988800048828124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,8,power_law_1.01,0.07889919877052307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,8,power_law_1.01,0.09096320271492005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,8,power_law_1.01,0.08088319897651672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,8,balanced,1.6189546585083008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,8,power_law_1.01,0.09041280150413514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,8,power_law_1.01,0.08435840010643006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,8,power_law_1.01,0.09200000166893005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,8,power_law_1.01,0.08424959778785705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,8,power_law_1.01,0.09392639994621277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,8,power_law_1.01,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,8,power_law_1.01,0.09427199959754944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,8,power_law_1.01,0.08888319730758668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,8,power_law_1.01,0.09564160108566284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,8,power_law_1.01,0.09350399971008301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,8,power_law_1.01,0.10253440141677857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,8,power_law_1.01,0.10035200119018554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,8,power_law_1.01,0.10500479936599731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,8,power_law_1.01,0.10831999778747559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,8,power_law_1.01,0.11119999885559081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,8,power_law_1.01,0.11859840154647827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,8,power_law_1.01,0.12015360593795776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,8,power_law_1.01,0.1357375979423523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,8,power_law_1.01,0.1331007957458496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,8,power_law_1.01,0.16440320014953613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,8,power_law_1.01,0.15249919891357422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,8,power_law_1.01,0.1931839942932129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,8,power_law_1.01,0.17038079500198364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,8,power_law_1.01,0.24647679328918456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,8,power_law_1.01,0.211027193069458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,8,power_law_1.01,0.2915776014328003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,8,balanced,2.264373302459717
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,8,power_law_1.01,0.24769279956817628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,8,power_law_1.01,0.4174975872039795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,8,power_law_1.01,0.32977919578552245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,8,power_law_1.01,0.52674560546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,8,power_law_1.01,0.39562880992889404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,8,power_law_1.01,0.7772543907165528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,8,power_law_1.01,0.587007999420166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,8,power_law_1.01,1.0275263786315918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,8,power_law_1.01,0.7115071773529053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,8,power_law_1.01,1.2392000198364257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,8,power_law_1.01,0.9263423919677735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,8,power_law_1.01,1.8589183807373046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,8,power_law_1.01,1.4333375930786132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,8,power_law_1.01,3.751500701904297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,8,power_law_1.01,2.855135917663574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,8,balanced,3.187562624613444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,8,power_law_1.2,0.0692031979560852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,8,power_law_1.2,0.07640960216522216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,8,power_law_1.2,0.06125440001487732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,8,power_law_1.2,0.06474879980087281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,8,power_law_1.2,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,8,power_law_1.2,0.08132479786872863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,8,power_law_1.2,0.08581759929656982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,8,power_law_1.2,0.08982399702072144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,8,power_law_1.2,0.09038720130920411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,8,power_law_1.2,0.06394240260124207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,8,power_law_1.2,0.09189119935035706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,128,balanced,0.03202133377393087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,8,power_law_1.2,0.07375360131263733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,128,balanced,0.032816000282764435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,8,power_law_1.2,0.0933568000793457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,8,power_law_1.2,0.05438719987869263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,128,balanced,0.033301333586374916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,8,power_law_1.2,0.0958079993724823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,128,balanced,0.03496533383925756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,8,power_law_1.2,0.06119679808616638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,8,power_law_1.2,0.0961023986339569
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,128,balanced,0.03497066597143809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,8,power_law_1.2,0.06544640064239501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,128,balanced,0.034832000732421875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,8,power_law_1.2,0.0985472023487091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,128,balanced,0.03676266719897588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,8,power_law_1.2,0.07219200134277344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,8,power_law_1.2,0.10496640205383301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,128,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,8,power_law_1.2,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,8,power_law_1.2,0.1069375991821289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,128,balanced,0.036202666660149894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,8,power_law_1.2,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,8,power_law_1.2,0.1140544056892395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,128,balanced,0.03630933413902918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,128,balanced,0.032085334261258446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,128,balanced,0.03187733391920725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,8,power_law_1.2,0.07948799729347229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,128,balanced,0.038021333515644073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,8,power_law_1.2,0.12284159660339355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,128,balanced,0.03209066639343897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,128,balanced,0.036789332826932274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,8,power_law_1.2,0.08086400032043457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,8,power_law_1.2,0.1365056037902832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,128,balanced,0.034789333740870156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,128,balanced,0.03881600002447764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,8,power_law_1.2,0.08354560136795045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,128,balanced,0.03624533365170161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,8,power_law_1.2,0.15460480451583863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,128,balanced,0.03799466788768768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,128,balanced,0.03640533238649368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,8,power_law_1.2,0.08298239707946778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,8,power_law_1.2,0.1719231963157654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,128,balanced,0.036570665736993156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,128,balanced,0.04046933352947235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,8,power_law_1.2,0.08804479837417603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,8,power_law_1.2,0.22077438831329346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,128,balanced,0.03572266548871994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,128,balanced,0.04212800165017446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,128,balanced,0.03643200049797694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,8,power_law_1.2,0.08609920144081115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,8,power_law_1.2,0.2694272041320801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,128,balanced,0.03640533238649368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,128,balanced,0.040896000961462654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,8,power_law_1.2,0.0956928014755249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,8,power_law_1.2,0.34647040367126464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,128,balanced,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,8,power_law_1.2,0.0985152006149292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,8,power_law_1.2,0.46638078689575196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,128,balanced,0.04258666435877482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,128,balanced,0.036576000352700554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,8,power_law_1.2,0.10438400506973267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,128,balanced,0.03999999910593033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,8,power_law_1.2,0.658784008026123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,128,balanced,0.04244266450405121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,8,power_law_1.2,0.12314879894256592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,128,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,8,power_law_1.2,0.8559552192687988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,128,balanced,0.0405173326532046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,8,power_law_1.2,0.13859200477600098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,8,power_law_1.2,1.0208000183105468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,128,balanced,0.04514666895071665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,128,balanced,0.043280000487963356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,8,power_law_1.2,0.1692479968070984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,8,power_law_1.2,1.512441635131836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,128,balanced,0.042634665966033936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,8,power_law_1.2,0.19776639938354493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,128,balanced,0.046911999583244324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,8,power_law_1.2,0.2636415958404541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,128,balanced,0.04470400015513102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,8,power_law_1.2,3.392736053466797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,8,power_law_1.2,0.3299263954162598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,128,balanced,0.04461866617202759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,8,power_law_1.2,0.4344319820404053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,128,balanced,0.048901334404945374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,128,balanced,0.05049600203831991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,8,power_law_1.2,0.5552063941955566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,8,power_law_1.2,0.8467328071594238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,128,balanced,0.05035733183224996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,8,power_law_1.2,0.9982272148132324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,128,balanced,0.051269332567850746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,8,power_law_1.2,1.384716796875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,128,balanced,0.05596800148487091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,8,power_law_1.2,2.085312080383301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,8,power_law_1.2,4.166227340698242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,128,balanced,0.05829333265622457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,128,balanced,0.05754133562246958
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,128,balanced,0.07499733567237854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,128,balanced,0.06474666794141133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,128,balanced,0.09462933739026387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,128,balanced,0.1346560021241506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,128,balanced,0.07611200213432312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,128,balanced,0.1648906668027242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,128,balanced,0.0921493371327718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,128,balanced,0.1962613264719645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,128,balanced,0.10831999778747559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,128,balanced,0.2857919931411743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,128,balanced,0.14457066853841147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,128,balanced,0.033520000676314034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,128,balanced,0.03373866776625315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,128,balanced,0.03465600063403448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,128,balanced,0.03640533238649368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,128,balanced,0.03499199946721395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,128,balanced,0.03637866675853729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,128,balanced,0.03774933268626531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,128,balanced,0.03699733316898346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,128,balanced,0.036544000109036766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,128,balanced,0.03842133283615112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,128,balanced,0.03806933263937632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,128,balanced,0.017637333522240322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,128,balanced,0.038912000755469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,128,balanced,0.02006400004029274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,128,balanced,0.038319999972979225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,128,balanced,0.020351999749739964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,128,balanced,0.0391893337170283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,128,balanced,0.021840001145998638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,128,balanced,0.04249600072701772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,128,balanced,0.032858667274316154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,128,balanced,0.02779199928045273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,128,balanced,0.04266666869322459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,128,balanced,0.028192001084486645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,128,balanced,0.04254400233427683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,128,balanced,0.02734400083621343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,128,balanced,0.028309332827727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,128,balanced,0.04462933540344238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,128,balanced,0.028037334481875103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,128,balanced,0.533456007639567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,128,balanced,0.02810666710138321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,128,balanced,0.04674133161703745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,128,balanced,0.028597332537174225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,128,balanced,0.02991466720898946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,128,balanced,0.04715733230113983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,128,balanced,0.03027733415365219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,128,balanced,0.032485333581765495
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,128,balanced,0.05036800106366476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,128,balanced,0.03150933235883713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,128,balanced,0.03251733382542928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,128,balanced,0.05228800078233083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,128,balanced,0.03425066669782003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,128,balanced,0.03653866549332937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,128,balanced,0.058287998040517174
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,128,balanced,0.038704000413417816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,128,balanced,0.04038933416207632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,128,balanced,0.0687360018491745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,128,balanced,0.046122665206591286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,128,balanced,0.05236800014972687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,128,balanced,0.07896533111731212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,128,balanced,0.2595946590105693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,128,balanced,0.06478400031725566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,128,balanced,0.10356266299883525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,128,balanced,0.08208000163237254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,128,balanced,0.113237331310908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,128,balanced,0.1227946678797404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,128,balanced,0.1406826674938202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,128,balanced,0.14493866761525473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,128,balanced,0.1689280072848002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,128,balanced,0.20191999276479086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,128,balanced,0.25410133600234985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,128,power_law_1.01,0.030508801341056824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,128,power_law_1.01,0.03167999982833862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,128,power_law_1.01,0.0308351993560791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,128,power_law_1.01,0.032876798510551454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,128,power_law_1.01,0.03343999981880188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,128,power_law_1.01,0.034246399998664856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,128,power_law_1.01,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,128,power_law_1.01,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,128,power_law_1.01,0.034329599142074584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,128,power_law_1.01,0.03480319976806641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,128,power_law_1.01,0.0356799989938736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,128,power_law_1.01,0.036320000886917114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,128,power_law_1.01,0.0359935998916626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,128,power_law_1.01,0.036550399661064145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,128,power_law_1.01,0.0390720009803772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,128,power_law_1.01,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,128,power_law_1.01,0.03976959884166718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,128,power_law_1.01,0.04094719886779785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,128,power_law_1.01,0.04288640022277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,128,power_law_1.01,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,128,power_law_1.01,0.04901759922504425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,128,power_law_1.01,0.054579198360443115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,128,balanced,0.3638240098953247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,128,power_law_1.01,0.029452800750732422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,128,power_law_1.01,0.061868798732757566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,128,power_law_1.01,0.030752000212669373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,128,power_law_1.01,0.0802944004535675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,128,power_law_1.01,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,128,power_law_1.01,0.09639040231704712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,128,power_law_1.01,0.031763198971748355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,128,power_law_1.01,0.13411200046539307
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,128,power_law_1.01,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,128,power_law_1.01,0.16370559930801393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,128,power_law_1.01,0.03504000008106232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,128,power_law_1.01,0.1947711944580078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,128,power_law_1.01,0.033228799700737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,128,power_law_1.01,0.03436160087585449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,128,power_law_1.01,0.3292736053466797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,128,power_law_1.01,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,128,power_law_1.01,0.5501376152038574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,128,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,128,power_law_1.01,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,128,power_law_1.01,0.03586559891700745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,128,power_law_1.01,0.03656960129737854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,128,balanced,0.47993067900339764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,128,power_law_1.01,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,128,power_law_1.01,0.0416703999042511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,128,power_law_1.01,0.039743998646736146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,128,power_law_1.01,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,128,power_law_1.01,0.04382719993591309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,128,power_law_1.01,0.045491200685501096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,128,power_law_1.01,0.050374400615692136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,128,power_law_1.01,0.05444480180740356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,128,power_law_1.01,0.06428160071372986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,128,power_law_1.01,0.07422080039978027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,128,power_law_1.01,0.10132479667663574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,128,power_law_1.01,0.13519359827041627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,128,power_law_1.01,0.18031359910964967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,128,power_law_1.01,0.23779840469360353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,128,power_law_1.01,0.2874367952346802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,128,power_law_1.01,0.4379583835601807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,128,power_law_1.01,0.9957247734069824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,128,power_law_1.01,0.031667199730873105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,128,power_law_1.01,0.03188480138778686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,128,power_law_1.01,0.031763198971748355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,128,power_law_1.01,0.033907198905944826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,128,power_law_1.01,0.03377920091152191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,128,power_law_1.01,0.03549439907073974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,128,power_law_1.01,0.03557760119438171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,128,power_law_1.01,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,128,power_law_1.01,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,128,power_law_1.01,0.0359935998916626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,128,power_law_1.01,0.03701759874820709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,128,power_law_1.01,0.016735999286174773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,128,power_law_1.2,0.03079040050506592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,128,power_law_1.01,0.03743360042572021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,128,power_law_1.2,0.03153280019760132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,128,power_law_1.01,0.018080000579357148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,128,power_law_1.01,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,128,power_law_1.2,0.031839999556541446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,128,power_law_1.01,0.01765120029449463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,128,power_law_1.01,0.03787519931793213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,128,power_law_1.2,0.03335680067539215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,128,power_law_1.01,0.019673599302768706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,128,power_law_1.01,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,128,power_law_1.2,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,128,power_law_1.01,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,128,power_law_1.2,0.034841600060462954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,128,power_law_1.01,0.04068480134010315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,128,power_law_1.01,0.024934400618076325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,128,power_law_1.2,0.03444480001926422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,128,power_law_1.01,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,128,power_law_1.01,0.025254398584365845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,128,power_law_1.2,0.034457600116729735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,128,power_law_1.01,0.04410879909992218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,128,power_law_1.01,0.024934400618076325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,128,power_law_1.2,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,128,power_law_1.01,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,128,power_law_1.01,0.025990399718284606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,128,power_law_1.2,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,128,power_law_1.01,0.05050879716873169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,128,power_law_1.01,0.026035198569297792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,128,power_law_1.2,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,128,power_law_1.01,0.0553600013256073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,128,power_law_1.01,0.025510400533676147
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,128,power_law_1.2,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,128,power_law_1.01,0.02688640058040619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,128,power_law_1.01,0.06350079774856568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,128,power_law_1.2,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,128,power_law_1.01,0.026956799626350402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,128,power_law_1.01,0.07399680018424988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,128,power_law_1.2,0.03696640133857727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,128,power_law_1.01,0.027871999144554137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,128,power_law_1.01,0.09655680060386658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,128,power_law_1.2,0.039603200554847715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,128,power_law_1.01,0.028543999791145323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,128,power_law_1.01,0.11400959491729737
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,128,power_law_1.2,0.0403903990983963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,128,power_law_1.01,0.029497599601745604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,128,power_law_1.01,0.15352319478988646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,128,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,128,power_law_1.01,0.030406400561332703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,128,power_law_1.01,0.19898879528045654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,128,power_law_1.2,0.042444801330566405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,128,power_law_1.01,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,128,power_law_1.01,0.22933759689331054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,128,power_law_1.2,0.04447360038757324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,128,power_law_1.01,0.035046398639678955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,128,power_law_1.01,0.3765887975692749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,128,power_law_1.2,0.04857600033283234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,128,power_law_1.2,0.029209598898887634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,128,power_law_1.01,0.03808639943599701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,128,power_law_1.2,0.054579198360443115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,128,power_law_1.2,0.030675199627876282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,128,power_law_1.01,0.7135871887207031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,128,power_law_1.01,0.03966079950332642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,128,power_law_1.2,0.06227200031280518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,128,power_law_1.2,0.031251201033592226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,128,power_law_1.01,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,128,power_law_1.2,0.0793727993965149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,128,power_law_1.2,0.032793599367141726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,128,power_law_1.01,0.05198720097541809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,128,power_law_1.2,0.10231679677963257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,128,power_law_1.2,0.033024001121521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,128,power_law_1.01,0.06445440053939819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,128,power_law_1.2,0.034246399998664856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,128,power_law_1.2,0.12021759748458863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,128,power_law_1.01,0.08279680013656616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,128,power_law_1.2,0.03460479974746704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,128,power_law_1.2,0.1697983980178833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,128,power_law_1.01,0.10837119817733765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,128,power_law_1.2,0.03557760119438171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,128,power_law_1.2,0.22351360321044922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,128,power_law_1.01,0.13482240438461304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,128,power_law_1.2,0.03497599959373474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,128,power_law_1.2,0.27123839855194093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,128,power_law_1.01,0.16277120113372803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,128,power_law_1.2,0.037785598635673524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,128,power_law_1.2,0.4024191856384277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,128,power_law_1.2,0.03559040129184723
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,128,power_law_1.01,0.24442880153656005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,128,power_law_1.2,0.03546879887580871
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,128,power_law_1.2,0.7747712135314941
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,128,power_law_1.01,0.46753921508789065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,128,power_law_1.2,0.03704319894313812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,128,power_law_1.2,0.03918719887733459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,128,power_law_1.2,0.04084480106830597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,128,power_law_1.2,0.041657599806785586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,128,power_law_1.2,0.04259839951992035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,128,power_law_1.2,0.04508799910545349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,128,power_law_1.2,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,128,power_law_1.2,0.0542464017868042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,128,power_law_1.2,0.0605184018611908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,128,power_law_1.2,0.07341439723968506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,128,power_law_1.2,0.09312000274658203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,128,power_law_1.2,0.1304255962371826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,128,power_law_1.2,0.1646399974822998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,128,power_law_1.2,0.2317888021469116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,128,power_law_1.2,0.29377920627593995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,128,power_law_1.2,0.36868479251861574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,128,power_law_1.2,0.5776768207550049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,128,power_law_1.2,1.1424192428588866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,128,power_law_1.2,0.03131519854068756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,128,power_law_1.2,0.03210879862308502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,128,power_law_1.2,0.03131519854068756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,128,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,128,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,128,power_law_1.2,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,128,power_law_1.2,0.03543039858341217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,128,power_law_1.2,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,128,power_law_1.2,0.03591679930686951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,128,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,128,power_law_1.2,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,128,power_law_1.2,0.017151999473571777
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,128,power_law_1.2,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,128,power_law_1.2,0.017504000663757326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,128,power_law_1.2,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,64,balanced,0.03178133318821589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,128,power_law_1.2,0.018265600502490997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,128,power_law_1.2,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,64,balanced,0.03262399882078171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,64,balanced,0.03211733450492223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,128,power_law_1.2,0.019801600277423857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,128,power_law_1.2,0.04083200097084046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,64,balanced,0.0346666673819224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,128,power_law_1.2,0.03167999982833862
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,128,power_law_1.2,0.040166398882865904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,64,balanced,0.03633599976698557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,128,power_law_1.2,0.02487040013074875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,128,power_law_1.2,0.04310399889945984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,64,balanced,0.038362666964530945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,128,power_law_1.2,0.026528000831604004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,128,power_law_1.2,0.04542079865932465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,64,balanced,0.03629333277543386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,64,balanced,0.03602666656176249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,128,power_law_1.2,0.026182401180267333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,128,power_law_1.2,0.04916479885578155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,64,balanced,0.03851733356714249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,128,power_law_1.2,0.025382399559020996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,128,power_law_1.2,0.05482879877090454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,64,balanced,0.037647999823093414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,128,power_law_1.2,0.025459200143814087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,128,power_law_1.2,0.06027519702911377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,64,balanced,0.036890665690104164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,128,power_law_1.2,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,64,balanced,0.03842133283615112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,128,power_law_1.2,0.0865664005279541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,64,balanced,0.03807999938726425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,128,power_law_1.2,0.11735039949417114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,64,balanced,0.038618666430314384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,128,power_law_1.2,0.14058239459991456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,64,balanced,0.042394667863845825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,128,power_law_1.2,0.19890559911727906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,64,balanced,0.04062933226426443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,128,power_law_1.2,0.2537856101989746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,64,balanced,0.04218666752179464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,64,balanced,0.031162666777769726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,128,power_law_1.2,0.3083904027938843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,64,balanced,0.043061330914497375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,64,balanced,0.03364266703526179
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,128,power_law_1.2,0.4707520008087158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,64,balanced,0.032602667808532715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,64,balanced,0.04394133388996124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,64,balanced,0.03510933369398117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,128,power_law_1.2,0.928006362915039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,64,balanced,0.034858666360378265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,64,balanced,0.036357333262761436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,64,balanced,0.04534400006135305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,64,balanced,0.03899733225504557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,64,balanced,0.03673599908749262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,64,balanced,0.04653333127498627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,64,balanced,0.036714665591716766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,64,balanced,0.03834133346875509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,64,balanced,0.038618666430314384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,64,balanced,0.05013866722583771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,64,balanced,0.04044266790151596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,64,balanced,0.03843733419974645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,64,balanced,0.040149333576361336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,64,balanced,0.052149335543314614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,64,balanced,0.04321066538492838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,64,balanced,0.04252266883850098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,64,balanced,0.042730664213498436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,64,balanced,0.04502399762471517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,64,balanced,0.059994667768478394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,64,balanced,0.046666666865348816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,64,balanced,0.04760533571243286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,64,balanced,0.06467733283837636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,64,balanced,0.05253866811593374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,64,balanced,0.056874667604764305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,64,balanced,0.08496000369389851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,64,balanced,0.06274666885534923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,64,balanced,0.07915199796358745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,64,balanced,0.09777599573135376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,64,balanced,0.09938666224479675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,64,balanced,0.14190399646759033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,64,balanced,0.11562666296958923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,64,balanced,0.17299199104309082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,64,balanced,0.15877866744995117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,64,balanced,0.20850666364034018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,64,power_law_1.01,0.03193599879741669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,64,power_law_1.01,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,64,power_law_1.01,0.031513598561286923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,64,balanced,0.304149329662323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,64,power_law_1.01,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,64,power_law_1.01,0.033657601475715636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,64,power_law_1.01,0.03516800105571747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,64,power_law_1.01,0.03484799861907959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,128,power_law_1.2,0.025363200902938844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,64,power_law_1.01,0.03439359962940216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,128,power_law_1.2,0.026035198569297792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,64,power_law_1.01,0.03711999952793121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,128,power_law_1.2,0.027673599123954774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,64,power_law_1.01,0.0350847989320755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,128,power_law_1.2,0.027935999631881713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,64,power_law_1.01,0.035776001214981076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,128,power_law_1.2,0.028940799832344054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,64,power_law_1.01,0.03607679903507233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,128,power_law_1.2,0.029766398668289184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,64,power_law_1.01,0.03644160032272339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,64,balanced,0.2821066578229268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,128,power_law_1.2,0.031071999669075014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,64,power_law_1.01,0.03727999925613403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,64,power_law_1.01,0.040940800309181215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,128,power_law_1.2,0.03223679959774017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,64,power_law_1.01,0.04108160138130188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,128,power_law_1.2,0.03487359881401062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,64,power_law_1.01,0.04100480079650879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,128,power_law_1.2,0.03767040073871612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,64,power_law_1.01,0.042694398760795595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,128,power_law_1.2,0.03927040100097656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,64,power_law_1.01,0.043609601259231565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,128,power_law_1.2,0.04534400105476379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,64,power_law_1.01,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,128,power_law_1.2,0.050367999076843264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,64,power_law_1.01,0.05047680139541626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,128,power_law_1.2,0.06544640064239501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,64,power_law_1.01,0.05570560097694397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,128,power_law_1.2,0.08391680121421814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,64,power_law_1.01,0.0643392026424408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,128,power_law_1.2,0.11165440082550049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,64,power_law_1.01,0.08174719810485839
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,128,power_law_1.2,0.14012800455093383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,64,power_law_1.01,0.09228799939155578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,128,power_law_1.2,0.16773760318756104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,64,power_law_1.01,0.13109760284423827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,128,power_law_1.2,0.25197439193725585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,64,power_law_1.01,0.15712000131607057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,128,power_law_1.2,0.47770237922668457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,64,power_law_1.01,0.19759360551834107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,64,power_law_1.01,0.2983871936798096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,64,power_law_1.01,0.5595839977264404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,64,balanced,0.565392017364502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,64,power_law_1.01,0.03107840120792389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,64,power_law_1.01,0.03089280128479004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,64,power_law_1.01,0.030527999997138976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,64,power_law_1.01,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,64,power_law_1.01,0.03457280099391937
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,64,power_law_1.01,0.035097599029541016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,64,power_law_1.01,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,64,power_law_1.01,0.035071998834609985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,64,power_law_1.01,0.036313599348068236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,64,power_law_1.01,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,64,power_law_1.01,0.037011200189590455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,64,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,64,power_law_1.01,0.03919999897480011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,64,power_law_1.01,0.038764798641204835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,64,power_law_1.01,0.04099839925765991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,64,power_law_1.01,0.0436928004026413
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,64,power_law_1.01,0.04317440092563629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,64,power_law_1.2,0.03185920119285583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,64,power_law_1.01,0.04562560021877289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,64,power_law_1.2,0.030649599432945252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,64,power_law_1.01,0.047251200675964354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,64,power_law_1.2,0.03108479976654053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,64,power_law_1.2,0.030956798791885377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,64,power_law_1.01,0.05132160186767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,64,power_law_1.2,0.03322240114212036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,64,power_law_1.2,0.030771198868751525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,64,power_law_1.01,0.05731199979782105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,64,power_law_1.2,0.03258880078792572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,64,power_law_1.2,0.031379199028015135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,64,power_law_1.01,0.06568319797515869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,64,power_law_1.2,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,64,power_law_1.2,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,64,power_law_1.01,0.0784704029560089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,64,power_law_1.2,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,64,power_law_1.2,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,64,power_law_1.01,0.11202559471130372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,64,power_law_1.2,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,64,power_law_1.2,0.03518719971179962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,64,power_law_1.01,0.14810880422592163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,64,power_law_1.2,0.03513599932193756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,64,power_law_1.2,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,64,power_law_1.01,0.20129919052124023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,64,power_law_1.2,0.03570559918880463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,64,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,64,power_law_1.01,0.2766592025756836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,64,power_law_1.2,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,64,power_law_1.2,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,64,power_law_1.01,0.32833280563354494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,128,balanced,0.04020266731580099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,64,power_law_1.2,0.036447998881340024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,64,power_law_1.2,0.03682560026645661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,64,power_law_1.01,0.5037759780883789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,128,balanced,0.040933333337306976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,64,power_law_1.2,0.03702400028705597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,64,power_law_1.2,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,128,balanced,0.04138666639725367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,64,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,64,power_law_1.01,1.0050432205200195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,64,power_law_1.2,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,128,balanced,0.04364799956480662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,64,power_law_1.2,0.04044800102710724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,64,power_law_1.2,0.038227200508117676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,128,balanced,0.042821332812309265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,64,power_law_1.2,0.04085760116577149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,64,power_law_1.2,0.03790720105171204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,128,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,64,power_law_1.2,0.04021120071411133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,64,power_law_1.2,0.04238080084323883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,128,balanced,0.04489600161711375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,64,power_law_1.2,0.04382719993591309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,64,power_law_1.2,0.04325119853019714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,128,balanced,0.04509866734345754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,64,power_law_1.2,0.044607999920845035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,64,power_law_1.2,0.04343680143356323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,128,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,64,power_law_1.2,0.047865599393844604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,64,power_law_1.2,0.047814399003982544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,128,balanced,0.044853334625562034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,64,power_law_1.2,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,64,power_law_1.2,0.050483202934265135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,128,balanced,0.044981335600217186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,64,power_law_1.2,0.06427519917488098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,64,power_law_1.2,0.05683839917182922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,128,balanced,0.0450133333603541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,64,power_law_1.2,0.07717120051383972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,64,power_law_1.2,0.06239359974861145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,128,balanced,0.04655999938646952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,64,power_law_1.2,0.09522560238838196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,64,power_law_1.2,0.07613440155982971
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,128,balanced,0.04609066744645437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,64,power_law_1.2,0.14559999704360962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,64,power_law_1.2,0.10380159616470337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,64,power_law_1.2,0.1800704002380371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,128,balanced,0.04667200148105621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,64,power_law_1.2,0.14133119583129883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,64,power_law_1.2,0.25372159481048584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,64,power_law_1.2,0.19102079868316652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,128,balanced,0.04853333532810211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,64,power_law_1.2,0.26044158935546874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,64,power_law_1.2,0.31023359298706055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,128,balanced,0.04938133557637533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,64,power_law_1.2,0.34602880477905273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,64,power_law_1.2,0.4730815887451172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,128,balanced,0.05034666756788889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,64,power_law_1.2,0.41276159286499026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,64,power_law_1.2,1.0845631599426269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,64,power_law_1.2,0.7780159950256348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,128,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,64,power_law_1.2,1.5628992080688477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,128,balanced,0.054048001766204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,128,balanced,0.056426664193471275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,128,balanced,0.05913599828879038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,128,balanced,0.06665066878000896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,128,balanced,0.07981866598129272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,128,balanced,0.09328533212343852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,128,balanced,0.11659733454386394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,128,balanced,0.13826666275660196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,128,balanced,0.1609653333822886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,128,balanced,0.03588266670703888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,128,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,128,balanced,0.03644266724586487
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,128,balanced,0.03827733298142751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,128,balanced,0.03884266565243403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,128,balanced,0.040336000422636666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,128,balanced,0.038880000511805214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,128,balanced,0.04085333396991094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,128,balanced,0.042405332128206887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,128,balanced,0.04098133246103922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,128,balanced,0.04295999805132548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,128,balanced,0.04058133314053217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,128,balanced,0.044682666659355164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,128,balanced,0.04457599918047587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,128,balanced,0.04670399924119314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,128,balanced,0.2357920010884603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,128,balanced,0.048714667558670044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,128,balanced,0.03604800005753835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,128,balanced,0.04853333532810211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,128,balanced,0.03724266588687897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,128,balanced,0.05074666440486908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,128,balanced,0.03673599908749262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,128,balanced,0.038618666430314384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,128,balanced,0.05250666538874308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,128,balanced,0.04029866556326548
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,128,balanced,0.04070399949947993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,128,balanced,0.019850666324297588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,128,balanced,0.04040000090996424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,128,balanced,0.05889600018660227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,128,balanced,0.020293333878119785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,128,balanced,0.04168533285458883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,128,balanced,0.022405333817005157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,128,balanced,0.04045333216587702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,128,balanced,0.024293333292007446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,128,balanced,0.04028266668319702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,128,balanced,0.06288533409436543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,128,balanced,0.0402453343073527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,128,balanced,0.040778666734695435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,128,balanced,0.03294399877389272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,128,balanced,0.04241600135962168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,128,balanced,0.03331733246644338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,128,balanced,0.04247466723124186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,128,balanced,0.032831999162832894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,128,balanced,0.07353066901365916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,128,balanced,0.042549331982930504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,128,balanced,0.03346133232116699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,128,balanced,0.044954667488733925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,128,balanced,0.033717334270477295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,128,balanced,0.034330666065216064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,128,balanced,0.04628799855709076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,128,balanced,0.03387200087308884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,128,balanced,0.04669333497683207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,128,balanced,0.09364266196886699
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,128,balanced,0.03374933451414108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,128,balanced,0.04879466692606608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,128,balanced,0.034490667283535004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,128,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,128,balanced,0.050666665037473045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,128,balanced,0.036943999429543815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,128,balanced,0.056703999638557434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,128,balanced,0.03836799909671148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,128,balanced,0.13462932904561362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,128,balanced,0.040063999593257904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,128,balanced,0.0581279993057251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,128,balanced,0.042352000872294106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,128,balanced,0.04743466774622599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,128,balanced,0.06910933554172516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,128,balanced,0.05247466762860616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,128,balanced,0.1614720026652018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,128,balanced,0.07719466586907704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,128,balanced,0.060965334375699363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,128,balanced,0.10058133800824483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,128,balanced,0.07938666641712189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,128,balanced,0.4148319959640503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,128,balanced,0.21780800819396973
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,128,balanced,0.1158026655515035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,128,balanced,0.1035093367099762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,128,balanced,0.12788266936937967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,128,balanced,0.15051199992497763
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,128,balanced,0.27342400948206586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,128,balanced,0.177456001440684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,128,balanced,0.1863200068473816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,128,balanced,0.22664000590642294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,128,balanced,0.3283626635869344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,128,balanced,0.2205280065536499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,128,balanced,0.2760213414827983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,128,balanced,0.32551467418670654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,128,balanced,0.49798401196797687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,128,balanced,0.4245706796646118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,128,balanced,0.6016639868418375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,128,power_law_1.01,0.038227200508117676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,128,power_law_1.01,0.039084801077842714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,128,power_law_1.01,0.0384768009185791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,128,power_law_1.01,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,128,power_law_1.01,0.041228801012039185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,128,balanced,0.9392159779866537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,128,power_law_1.01,0.04120959937572479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,128,power_law_1.01,0.041920000314712526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,128,power_law_1.01,0.042028799653053284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,128,power_law_1.01,0.042368000745773314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,128,power_law_1.01,0.04240640103816986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,128,power_law_1.01,0.04281600117683411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,128,balanced,0.8160906632741293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,128,power_law_1.01,0.04270080029964447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,128,power_law_1.01,0.04567039906978607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,128,power_law_1.01,0.04373759925365448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,128,power_law_1.01,0.04542720019817352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,128,power_law_1.01,0.04785279929637909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,128,power_law_1.01,0.04859519898891449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,128,power_law_1.01,0.050310397148132326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,128,power_law_1.01,0.0541055977344513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,128,power_law_1.01,0.06195840239524841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,128,power_law_1.01,0.06821759939193725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,128,power_law_1.01,0.08636159896850586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,128,power_law_1.01,0.10196479558944702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,128,power_law_1.01,0.1316799998283386
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,128,power_law_1.01,0.16819839477539061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,128,power_law_1.01,0.23356800079345702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,128,power_law_1.01,0.29868159294128416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,128,power_law_1.01,0.37811200618743895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,128,power_law_1.01,0.6292672157287598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,128,power_law_1.01,1.3192895889282226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,128,power_law_1.01,0.03462400138378143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,128,power_law_1.01,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,128,power_law_1.01,0.03601279854774475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,128,power_law_1.01,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,128,power_law_1.01,0.03754880130290985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,128,power_law_1.01,0.03443840146064758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,128,power_law_1.01,0.038464000821113585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,128,power_law_1.01,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,128,power_law_1.01,0.034841600060462954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,128,power_law_1.01,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,128,power_law_1.01,0.034841600060462954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,128,power_law_1.01,0.039078399538993835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,128,power_law_1.01,0.03633280098438263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,128,power_law_1.01,0.04013440012931824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,128,power_law_1.01,0.037555199861526486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,128,power_law_1.01,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,128,power_law_1.01,0.03783040046691895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,128,power_law_1.01,0.03996799886226654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,128,power_law_1.01,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,128,power_law_1.01,0.03868800103664398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,128,power_law_1.01,0.04089600145816803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,128,power_law_1.01,0.038924801349639895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,128,power_law_1.01,0.04222080111503601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,128,power_law_1.01,0.038380798697471616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,128,power_law_1.01,0.044863998889923096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,128,power_law_1.01,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,128,power_law_1.01,0.046323201060295104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,128,power_law_1.01,0.018643200397491455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,128,power_law_1.01,0.04056960046291351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,128,power_law_1.01,0.04751999974250794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,128,power_law_1.01,0.01955839991569519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,128,power_law_1.01,0.04142079949378967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,128,power_law_1.01,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,128,power_law_1.01,0.019519999623298645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,128,power_law_1.01,0.04085119962692261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,128,power_law_1.01,0.05505920052528381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,128,power_law_1.01,0.022771200537681578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,128,power_law_1.01,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,128,power_law_1.01,0.06401919722557067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,128,power_law_1.01,0.03842560052871704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,128,power_law_1.01,0.04576640129089356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,128,power_law_1.01,0.0738431990146637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,128,power_law_1.01,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,128,power_law_1.01,0.04677119851112366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,128,power_law_1.01,0.09477120041847228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,128,power_law_1.01,0.03121280074119568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,128,power_law_1.01,0.05093119740486145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,128,power_law_1.01,0.12467199563980103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,128,power_law_1.01,0.03162879943847656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.01,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,128,power_law_1.01,0.17596800327301027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,128,power_law_1.01,0.0320576012134552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.01,0.060844802856445314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,128,power_law_1.01,0.2197887897491455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,128,power_law_1.01,0.031539198756217954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.01,0.07022719979286193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,128,power_law_1.01,0.31217279434204104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,128,power_law_1.01,0.032262399792671204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.01,0.08257920145988465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,128,power_law_1.01,0.40869760513305664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,128,power_law_1.01,0.03285120129585266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.01,0.10305279493331909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,128,power_law_1.01,0.5106112003326416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,128,power_law_1.01,0.03271040022373199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.01,0.1334720015525818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,128,power_law_1.01,0.7637184143066407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,128,power_law_1.01,0.032767999172210696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.01,0.1641152024269104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,128,power_law_1.01,0.03364480137825012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,128,power_law_1.01,1.5275967597961426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.01,0.24316160678863524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,128,power_law_1.01,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.01,0.31439359188079835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,128,power_law_1.2,0.03808000087738037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,128,power_law_1.01,0.035648000240325925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,128,power_law_1.01,0.3891263961791992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,128,power_law_1.2,0.038464000821113585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,128,power_law_1.01,0.040006399154663086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,128,power_law_1.01,0.5635968208312988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,128,power_law_1.2,0.03927040100097656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.01,0.041254401206970215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,128,power_law_1.2,0.04010879993438721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,128,power_law_1.01,1.1399616241455077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.01,0.04628480076789856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,128,power_law_1.2,0.04166400134563446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.01,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,128,power_law_1.2,0.04256640076637268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.01,0.05978239774703979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,128,power_law_1.2,0.04267520010471344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.01,0.07683839797973632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,128,power_law_1.2,0.04226559996604919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.01,0.09869440197944641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,128,power_law_1.2,0.042656001448631284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.01,0.12522879838943482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,128,power_law_1.2,0.04229759871959686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.01,0.16565760374069213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,128,power_law_1.2,0.0429504007101059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,128,power_law_1.2,0.04309119880199432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.01,0.21928319931030274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,128,power_law_1.2,0.04337919950485229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,128,power_law_1.01,0.2638144016265869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,128,power_law_1.2,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,128,power_law_1.01,0.3993407964706421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,128,power_law_1.2,0.047219198942184445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,128,power_law_1.2,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,128,power_law_1.01,0.773689603805542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,128,power_law_1.2,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,128,power_law_1.2,0.054201602935791016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,128,power_law_1.2,0.05792639851570129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,128,power_law_1.2,0.06961920261383056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,128,power_law_1.2,0.08460800051689148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,128,power_law_1.2,0.106387197971344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,128,power_law_1.2,0.1273151993751526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,128,power_law_1.2,0.1689471960067749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,128,power_law_1.2,0.22049920558929442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,128,power_law_1.2,0.295961594581604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,128,power_law_1.2,0.42282881736755373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,128,power_law_1.2,0.5282944202423095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,128,power_law_1.2,0.8259967803955078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,128,power_law_1.2,1.6168127059936523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,128,power_law_1.2,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,128,power_law_1.2,0.035180801153182985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,128,power_law_1.2,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,128,power_law_1.2,0.036735999584198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,128,power_law_1.2,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,128,power_law_1.2,0.037567999958992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,128,power_law_1.2,0.036601600050926206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,128,power_law_1.2,0.03893760144710541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,128,power_law_1.2,0.03390080034732819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,128,power_law_1.2,0.03871360123157501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,128,power_law_1.2,0.037036800384521486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,128,power_law_1.2,0.039103999733924866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,128,power_law_1.2,0.03741439878940582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,128,power_law_1.2,0.039129599928855896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,128,power_law_1.2,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,128,power_law_1.2,0.03959679901599884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,128,power_law_1.2,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,128,power_law_1.2,0.04046080112457275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,128,power_law_1.2,0.03852159976959228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,128,power_law_1.2,0.04053759872913361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,128,power_law_1.2,0.03829759955406189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,128,power_law_1.2,0.04120959937572479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,128,power_law_1.2,0.03870719969272614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,128,power_law_1.2,0.042438399791717527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,128,power_law_1.2,0.039878401160240176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,128,power_law_1.2,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,128,power_law_1.2,0.047014400362968445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,128,power_law_1.2,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,128,power_law_1.2,0.05400959849357605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,128,power_law_1.2,0.05761920213699341
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,128,power_law_1.2,0.01852159947156906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,128,power_law_1.2,0.07023360133171082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,128,power_law_1.2,0.019270400702953338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,128,power_law_1.2,0.09077119827270508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,128,power_law_1.2,0.019648000597953796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,128,power_law_1.2,0.11870720386505126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,128,power_law_1.2,0.021804800629615782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,128,power_law_1.2,0.15380480289459228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,128,power_law_1.2,0.03912320137023926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,128,power_law_1.2,0.21274878978729247
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,128,power_law_1.2,0.030873599648475646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,128,power_law_1.2,0.2558784008026123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,128,power_law_1.2,0.031590399146080014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,128,power_law_1.2,0.3791615962982178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,128,power_law_1.2,0.030342400074005127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,128,power_law_1.2,0.5019008159637451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,128,power_law_1.2,0.031999999284744264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,128,power_law_1.2,0.6147200107574463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,128,power_law_1.2,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,128,power_law_1.2,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,128,power_law_1.2,0.991385555267334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,128,power_law_1.2,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,128,power_law_1.2,1.9345855712890625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,128,power_law_1.2,0.03253760039806366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,64,balanced,0.038165333370367684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,128,power_law_1.2,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,64,balanced,0.04195733368396759
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,64,balanced,0.03856533269087473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,128,power_law_1.2,0.034176000952720643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,64,balanced,0.041008000572522484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,128,power_law_1.2,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,64,balanced,0.04378133515516917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,128,power_law_1.2,0.03607679903507233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,64,balanced,0.043050666650136314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,64,balanced,0.044954667488733925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,128,power_law_1.2,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,64,balanced,0.045109331607818604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.2,0.041407999396324155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,64,balanced,0.04324266811211904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.2,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,64,balanced,0.04457066456476847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.2,0.050393599271774295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,64,balanced,0.04543466866016388
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.2,0.061299198865890504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,64,balanced,0.044581333796183266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,64,balanced,0.04600533346335093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.2,0.07956479787826538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,64,balanced,0.04452266792456309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.2,0.1021888017654419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,64,balanced,0.04667733112970988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.2,0.12630399465560913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,64,balanced,0.04850666721661886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.2,0.17568639516830445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.2,0.22663040161132814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,64,balanced,0.04836266736189524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,128,power_law_1.2,0.27151999473571775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,64,balanced,0.050997331738471985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,128,power_law_1.2,0.42281599044799806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,64,balanced,0.050330668687820435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,128,power_law_1.2,0.8141695976257324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,64,balanced,0.05531733234723409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,64,balanced,0.05676800012588501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,64,balanced,0.062037333846092224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,64,balanced,0.06858666737874348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,64,balanced,0.08504533767700195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,64,balanced,0.09898666540781657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,64,balanced,0.12235732873280843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,64,balanced,0.14834133783976236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,64,balanced,0.17294933398564658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,64,balanced,0.03578133384386698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,64,balanced,0.035802667339642845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,64,balanced,0.03486400097608566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,64,balanced,0.03722133239110311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,64,balanced,0.040591999888420105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,64,balanced,0.04131199916203817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,128,power_law_1.2,0.039647999405860904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,64,balanced,0.04093866546948751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,128,power_law_1.2,0.040889599919319154
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,64,balanced,0.04230933388074239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,128,power_law_1.2,0.041465601325035094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,64,balanced,0.04257066547870636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,128,power_law_1.2,0.045612800121307376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,64,balanced,0.04189866781234741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,64,balanced,0.2512960036595662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,128,power_law_1.2,0.04524799883365631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,64,balanced,0.04275733232498169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,128,power_law_1.2,0.048332801461219786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,64,balanced,0.04257600009441376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,64,balanced,0.04479999840259552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,128,power_law_1.2,0.052300798892974856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,64,balanced,0.04408533374468485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.2,0.05617280006408691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,64,balanced,0.046336000164349876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.2,0.06686720252037048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.2,0.07527040243148804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,64,balanced,0.0484799991051356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.2,0.09836159944534302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,64,balanced,0.04866133133570353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.2,0.1207360029220581
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,64,balanced,0.052245333790779114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.2,0.16668800115585328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,64,balanced,0.05266133447488149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.2,0.2082047939300537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.2,0.3063103914260864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,64,balanced,0.05923733115196228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.2,0.39661440849304197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,128,power_law_1.2,0.4842559814453125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,64,balanced,0.06252266466617584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,128,power_law_1.2,0.750271987915039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,64,balanced,0.0359253336985906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,128,power_law_1.2,1.4630144119262696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,64,balanced,0.03850133220354716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,64,balanced,0.07700799902280171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,64,balanced,0.036517334481080375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,64,balanced,0.03899200012286504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,64,balanced,0.04074666649103165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,64,balanced,0.041989331444104515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,64,balanced,0.043023998538653054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,64,balanced,0.09790933132171631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,64,balanced,0.042463997999827065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,64,balanced,0.04250133534272512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,64,balanced,0.0425546665986379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,64,balanced,0.043007999658584595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,64,balanced,0.04433600107828776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,64,balanced,0.1383840044339498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,64,balanced,0.043103997906049095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,64,balanced,0.043951998154322304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,64,balanced,0.048613334695498146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,64,balanced,0.048751999934514366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,64,balanced,0.16657599806785583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,64,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,64,balanced,0.05072533090909322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,64,balanced,0.05443733433882395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,64,balanced,0.05643733342488607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,64,balanced,0.4541706641515096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,64,balanced,0.06084799766540527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,64,balanced,0.22642133633295694
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,64,balanced,0.07284800211588542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,64,balanced,0.08182933429876964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,64,balanced,0.28355733553568524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,64,balanced,0.1053706705570221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,64,balanced,0.12231466174125671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,64,balanced,0.3402186632156372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,64,balanced,0.1613759994506836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,64,balanced,0.1991200049718221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,64,balanced,0.5174560149510702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,64,balanced,0.23911466201146445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,64,balanced,0.01931200052301089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,64,balanced,0.35069866975148517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,64,balanced,0.025775998830795288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,64,balanced,0.019978666057189304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,64,balanced,0.020319999506076176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,64,balanced,0.023919999599456787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,64,balanced,0.0405973345041275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,64,balanced,0.032399999598662056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,64,balanced,0.03253333270549774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,64,balanced,0.032629333436489105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,64,balanced,0.978869358698527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,64,balanced,0.03245333333810171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,64,balanced,0.032127998769283295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,64,balanced,0.03229333211978277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,64,balanced,0.03446933378775915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,64,balanced,0.034341332813103996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,64,balanced,0.037647999823093414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,64,balanced,0.0363520011305809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,64,power_law_1.01,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,64,balanced,0.03663466622432073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,64,power_law_1.01,0.03893760144710541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,64,power_law_1.01,0.03932799994945526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,64,balanced,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,64,power_law_1.01,0.041791999340057374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,64,balanced,0.04605866471926371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,64,power_law_1.01,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,64,power_law_1.01,0.04357120096683502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,64,power_law_1.01,0.04232960045337677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,64,balanced,0.0489279975493749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,64,power_law_1.01,0.04208639860153198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,64,power_law_1.01,0.04224640130996704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,64,balanced,0.05496533215045929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,64,power_law_1.01,0.04238080084323883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,64,power_law_1.01,0.043161600828170776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,64,power_law_1.01,0.04327679872512817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,64,balanced,0.06921066840489705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,64,power_law_1.01,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,64,power_law_1.01,0.04390400052070618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,64,power_law_1.01,0.047328001260757445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,64,power_law_1.01,0.04692479968070984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,64,power_law_1.01,0.04878720045089722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,64,balanced,0.08784533540407817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,64,power_law_1.01,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,64,power_law_1.01,0.05438079833984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,64,power_law_1.01,0.06104959845542908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,64,power_law_1.01,0.06757760047912598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,64,power_law_1.01,0.07994239926338195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,64,balanced,0.10989333192507426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,64,power_law_1.01,0.09035519957542419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,64,balanced,0.6505759954452515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,64,power_law_1.01,0.1320512056350708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,64,power_law_1.01,0.1711743950843811
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,64,power_law_1.01,0.24017279148101806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,64,power_law_1.01,0.2793920040130615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,64,power_law_1.01,0.351200008392334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,64,power_law_1.01,0.4798463821411133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,64,balanced,0.14086932937304178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,64,power_law_1.01,1.0547519683837892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,64,balanced,0.19299733638763428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,64,balanced,0.25041600068410236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,64,balanced,0.30294400453567505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,64,power_law_1.01,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,64,power_law_1.01,0.03523840010166168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,64,power_law_1.01,0.035174399614334106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,64,power_law_1.01,0.03716480135917664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,64,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,64,power_law_1.01,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,64,power_law_1.01,0.03991680145263672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,64,power_law_1.01,0.03996799886226654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,64,power_law_1.01,0.040300801396369934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,64,power_law_1.01,0.03966079950332642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,64,power_law_1.01,0.04158720076084137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,64,power_law_1.01,0.041433599591255185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,64,power_law_1.01,0.04270719885826111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,64,power_law_1.01,0.04216960072517395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,64,power_law_1.01,0.04593920111656189
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,64,power_law_1.01,0.04713599979877472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,64,power_law_1.01,0.04852479994297028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,64,power_law_1.01,0.05203199982643127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,64,power_law_1.01,0.05842559933662415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,64,balanced,0.47281066576639813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,64,power_law_1.01,0.06541439890861511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,64,power_law_1.01,0.07737600207328796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,64,power_law_1.01,0.10172799825668336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,64,power_law_1.01,0.12811520099639892
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,64,power_law_1.01,0.17745920419692993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,64,power_law_1.01,0.23235199451446534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,64,power_law_1.01,0.3032448053359985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,64,power_law_1.01,0.3746112108230591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,64,power_law_1.01,0.5086527824401855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,64,power_law_1.01,0.7628032207489014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,64,power_law_1.01,1.5175488471984864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,64,power_law_1.01,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,64,power_law_1.01,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,64,power_law_1.01,0.03672960102558136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,64,power_law_1.01,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,64,power_law_1.01,0.038515201210975646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,64,power_law_1.01,0.04020479917526245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,64,power_law_1.01,0.040863999724388124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,64,power_law_1.01,0.017958399653434754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,64,power_law_1.01,0.04105600118637085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,64,power_law_1.01,0.01895039975643158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,64,power_law_1.01,0.04097279906272888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,64,power_law_1.01,0.01804800033569336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,64,power_law_1.01,0.04030719995498657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,64,power_law_1.01,0.01825920045375824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,64,power_law_1.01,0.04175359904766083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,64,power_law_1.01,0.02260479927062988
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,64,power_law_1.01,0.04199039936065674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,64,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,64,power_law_1.01,0.0431551992893219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,64,power_law_1.01,0.031071999669075014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,64,power_law_1.01,0.04306559860706329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,64,power_law_1.01,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,64,power_law_1.01,0.04671359956264496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,64,power_law_1.01,0.0313728004693985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,64,power_law_1.01,0.047251200675964354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,64,power_law_1.01,0.03086720108985901
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,64,power_law_1.01,0.0526528000831604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,64,power_law_1.01,0.03244799971580505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,64,power_law_1.01,0.05370879769325256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,64,power_law_1.01,0.03253760039806366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.01,0.05800319910049438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,64,power_law_1.01,0.03196159899234772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.01,0.06362239718437195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,64,power_law_1.01,0.031641599535942075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.01,0.07336959838867188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,64,power_law_1.01,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.01,0.09194239974021912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,64,power_law_1.01,0.034246399998664856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.01,0.11139839887619019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,64,power_law_1.01,0.03583360016345978
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.01,0.15034879446029664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,64,power_law_1.01,0.0395583987236023
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.01,0.17512960433959962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,64,balanced,0.9152746995290121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.01,0.044556799530982974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.01,0.2473855972290039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.01,0.04888319969177246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.01,0.3275455951690674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.01,0.05273600220680237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,64,power_law_1.01,0.3889791965484619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.01,0.06414080262184144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,64,power_law_1.01,0.599513578414917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.01,0.08054400086402894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,64,power_law_1.01,1.2239551544189453
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.01,0.1044927954673767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.01,0.13133440017700196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.01,0.1730239987373352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.01,0.22271358966827393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,64,power_law_1.01,0.2749631881713867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,64,power_law_1.01,0.4432831764221191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,64,power_law_1.01,0.8548543930053711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,64,power_law_1.2,0.04012799859046936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,64,power_law_1.2,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,64,power_law_1.2,0.03738879859447479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,64,power_law_1.2,0.04022400081157684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,64,power_law_1.2,0.04167680144309997
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,64,power_law_1.2,0.04188160002231598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,64,power_law_1.2,0.0427839994430542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,64,power_law_1.2,0.043110400438308716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,64,power_law_1.2,0.04224640130996704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,64,power_law_1.2,0.04270719885826111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,64,power_law_1.2,0.04302079975605011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,64,power_law_1.2,0.043654400110244754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,64,power_law_1.2,0.04449920058250427
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,64,power_law_1.2,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,64,power_law_1.2,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,64,power_law_1.2,0.04888319969177246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,64,power_law_1.2,0.0500544011592865
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,64,power_law_1.2,0.0530239999294281
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,64,power_law_1.2,0.05719040036201477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,64,power_law_1.2,0.06392959952354431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,64,power_law_1.2,0.0732159972190857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,64,power_law_1.2,0.09256319999694824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,64,power_law_1.2,0.12055679559707641
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,64,power_law_1.2,0.14711040258407593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,64,power_law_1.2,0.21809279918670654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,64,power_law_1.2,0.034835198521614076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,64,power_law_1.2,0.32284159660339357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,64,power_law_1.2,0.035436800122261046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,64,power_law_1.2,0.42315521240234377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,64,power_law_1.2,0.033932799100875856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,64,power_law_1.2,0.5773568153381348
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,64,power_law_1.2,0.03680639863014221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,64,power_law_1.2,0.03813759982585907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,64,power_law_1.2,0.9451199531555176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,64,power_law_1.2,0.03937920033931732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,64,power_law_1.2,1.9968448638916017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,64,power_law_1.2,0.03937279880046844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,64,power_law_1.2,0.040550398826599124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,64,power_law_1.2,0.040268799662590025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,64,power_law_1.2,0.04090240001678467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,64,power_law_1.2,0.04168320000171662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,64,power_law_1.2,0.03772160112857818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,64,power_law_1.2,0.04261760115623474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,64,power_law_1.2,0.03741439878940582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,64,power_law_1.2,0.042508798837661746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,64,power_law_1.2,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,64,power_law_1.2,0.042752000689506534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,64,power_law_1.2,0.03877759873867035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,64,power_law_1.2,0.04742400050163269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,64,power_law_1.2,0.03944959938526153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,64,power_law_1.2,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,64,power_law_1.2,0.04092159867286682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,64,power_law_1.2,0.05090559720993042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,64,power_law_1.2,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,64,power_law_1.2,0.01817599982023239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,64,power_law_1.2,0.056518399715423585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,64,power_law_1.2,0.04026240110397339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,64,power_law_1.2,0.059308797121047974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,64,power_law_1.2,0.018892799317836762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,64,power_law_1.2,0.041094401478767396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,64,power_law_1.2,0.07296000123023987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,64,power_law_1.2,0.018918399512767792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,64,power_law_1.2,0.04188160002231598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,64,power_law_1.2,0.08747519850730896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,64,power_law_1.2,0.019340799748897554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,64,power_law_1.2,0.041305598616600034
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,64,power_law_1.2,0.11633919477462769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,64,power_law_1.2,0.021113599836826324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,64,power_law_1.2,0.04291200041770935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,64,power_law_1.2,0.1538432002067566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,64,power_law_1.2,0.03917439877986908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,64,power_law_1.2,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,64,power_law_1.2,0.20631680488586426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,64,power_law_1.2,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,64,power_law_1.2,0.042771199345588685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,64,power_law_1.2,0.2757375955581665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,64,power_law_1.2,0.047225600481033324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,64,power_law_1.2,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,64,power_law_1.2,0.47599358558654786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,64,power_law_1.2,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,64,power_law_1.2,0.031814399361610415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,64,power_law_1.2,0.5431551933288574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,64,power_law_1.2,0.05120000243186951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,64,power_law_1.2,0.031097599864006044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,64,power_law_1.2,0.722111988067627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,64,power_law_1.2,0.05596799850463867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,64,power_law_1.2,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,64,power_law_1.2,1.2246784210205077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.2,0.059545600414276124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,64,power_law_1.2,0.03284479975700379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.2,0.0697983980178833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,64,power_law_1.2,2.197427177429199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,64,power_law_1.2,0.03212159872055054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.2,0.08181759715080261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,64,power_law_1.2,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.2,0.10432000160217285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,64,power_law_1.2,0.03273600041866302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.2,0.13118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,64,power_law_1.2,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.2,0.16785919666290283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,64,power_law_1.2,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.2,0.26702721118927003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,64,power_law_1.2,0.0398719996213913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.2,0.3384063959121704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.2,0.04344319999217987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.2,0.4961599826812744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.2,0.048390400409698484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,64,power_law_1.2,0.4952127933502197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.2,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,64,power_law_1.2,0.9202495574951172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.2,0.0666815996170044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,64,power_law_1.2,1.8360448837280274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.2,0.08327680230140685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.2,0.10734720230102539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.2,0.1326591968536377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.2,0.18985600471496583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.2,0.24049921035766603
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,64,power_law_1.2,0.2950592041015625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,64,power_law_1.2,0.47185277938842773
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,64,power_law_1.2,0.9139519691467285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,32,balanced,0.03823466598987579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,32,balanced,0.036874666810035706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,32,balanced,0.03723733375469843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,32,balanced,0.040693332751592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,32,balanced,0.0440533310174942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,32,balanced,0.04611733555793762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,32,balanced,0.04494933287302653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,32,balanced,0.04289066791534424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,32,balanced,0.04496533175309499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,32,balanced,0.04457599918047587
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,32,balanced,0.044480000933011375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,32,balanced,0.04496000210444132
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,32,balanced,0.04465066889921824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,32,balanced,0.04654933512210846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,32,balanced,0.04901333153247833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,32,balanced,0.049216002225875854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,32,balanced,0.05075199902057648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,32,balanced,0.0506879985332489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,32,balanced,0.05285866558551788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,32,balanced,0.054986665646235146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,32,balanced,0.03417599946260452
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,32,balanced,0.03482133398453394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,32,balanced,0.05735999842484792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,32,balanced,0.03481066723664602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,32,balanced,0.03875733415285746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,32,balanced,0.0420959989229838
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,32,balanced,0.042437334855397545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,32,balanced,0.06723733246326447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,32,balanced,0.041050667564074196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,32,balanced,0.04311466713746389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,32,balanced,0.04306666553020477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,32,balanced,0.04268266757329305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,32,balanced,0.04418133199214935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,32,balanced,0.07333866755167644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,32,balanced,0.04299733539422353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,32,balanced,0.046944002310434975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,32,balanced,0.045040001471837364
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,32,balanced,0.046997333566347756
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,32,balanced,0.048485333720842995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,32,balanced,0.09370666742324829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,32,balanced,0.05083733300367991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,32,balanced,0.05277866621812185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,32,power_law_1.01,0.04073599874973297
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,32,power_law_1.01,0.04017919898033142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,32,balanced,0.05484800040721893
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,32,power_law_1.01,0.038438400626182555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,32,power_law_1.01,0.04039680063724518
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,32,balanced,0.06307200094064076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,32,power_law_1.01,0.04092159867286682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,32,power_law_1.01,0.04340479969978332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,32,balanced,0.10748799641927083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,32,power_law_1.01,0.0427839994430542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,32,balanced,0.06695466736952464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,32,power_law_1.01,0.04337919950485229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,32,power_law_1.01,0.043424001336097716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,32,power_law_1.01,0.03736959993839264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,32,balanced,0.0843999981880188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,32,power_law_1.01,0.04328320026397705
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,32,power_law_1.01,0.03589119911193848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,32,power_law_1.01,0.04459519982337952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,32,power_law_1.01,0.03456639945507049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,32,power_law_1.01,0.04431999921798706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,32,power_law_1.01,0.036883199214935304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,32,balanced,0.10355732838312785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,32,power_law_1.01,0.04447360038757324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,32,power_law_1.01,0.038700801134109494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,32,power_law_1.01,0.04636160135269165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,32,power_law_1.01,0.04015359878540039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,32,power_law_1.01,0.04849919974803925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,32,balanced,0.13850133617719015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,32,power_law_1.01,0.04102399945259094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,32,power_law_1.01,0.04912000000476837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,32,power_law_1.01,0.04115839898586273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,32,balanced,0.1463573376337687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,32,power_law_1.01,0.051072001457214355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,32,power_law_1.01,0.041465601325035094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,32,power_law_1.01,0.053401601314544675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,32,power_law_1.01,0.04163840115070343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,32,power_law_1.01,0.05623679757118225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,32,power_law_1.01,0.042335999011993405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,32,power_law_1.01,0.06706560254096985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,32,power_law_1.01,0.04239999949932098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,32,power_law_1.01,0.07200000286102295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,32,power_law_1.01,0.043699198961257936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,32,balanced,0.17749865849812826
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,32,power_law_1.01,0.0869055986404419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,32,power_law_1.01,0.044915199279785156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,32,power_law_1.01,0.047276800870895384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,32,power_law_1.01,0.10058239698410035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,32,power_law_1.01,0.049184000492095946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,32,power_law_1.01,0.12906240224838256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,32,power_law_1.01,0.04978559911251068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,32,power_law_1.01,0.1562559962272644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,32,power_law_1.01,0.05612159967422485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,32,power_law_1.01,0.24269440174102783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,32,power_law_1.01,0.059513598680496216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,32,balanced,0.1648960014184316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,32,power_law_1.01,0.2931328058242798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,32,power_law_1.01,0.06669440269470214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,32,power_law_1.01,0.3623872041702271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,32,balanced,0.24177066485087076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,32,power_law_1.01,0.0816703975200653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,32,power_law_1.01,0.6069503784179687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,32,power_law_1.01,0.10893440246582031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,32,power_law_1.01,0.13822720050811768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,32,power_law_1.01,1.028883171081543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,32,power_law_1.01,0.18845440149307252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,32,power_law_1.01,0.23467519283294677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,32,power_law_1.01,0.31480960845947265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,32,power_law_1.01,0.4560512065887451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,32,power_law_1.01,0.5242047786712647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,32,balanced,0.30589866638183594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,32,power_law_1.01,0.8291071891784668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,32,power_law_1.01,1.503065586090088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,32,balanced,0.195360004901886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,32,balanced,0.3670186599095662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,32,balanced,0.2869120041529338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,32,balanced,0.5576800107955933
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,32,balanced,1.0685760180155437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,32,balanced,0.5248746474583944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,32,power_law_1.2,0.041247999668121337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,32,power_law_1.2,0.037599998712539676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,32,power_law_1.2,0.037222400307655334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,32,power_law_1.2,0.04071680009365082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,32,power_law_1.2,0.041222399473190306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,32,power_law_1.2,0.043219199776649474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,32,power_law_1.2,0.04291200041770935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,32,power_law_1.2,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,32,power_law_1.2,0.04296959936618805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,32,power_law_1.2,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,32,power_law_1.2,0.04382719993591309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,32,power_law_1.2,0.0356799989938736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,32,power_law_1.2,0.04319359958171844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,32,power_law_1.2,0.037625598907470706
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,32,power_law_1.2,0.03770880103111267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,32,power_law_1.2,0.04423680007457733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,32,power_law_1.2,0.04074879884719849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,32,power_law_1.2,0.04527359902858734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,32,power_law_1.2,0.04059520065784454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,32,power_law_1.2,0.04450559914112091
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,32,power_law_1.2,0.04010240137577057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,32,power_law_1.2,0.04656639993190766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,32,power_law_1.2,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,32,power_law_1.2,0.04981119930744171
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,32,power_law_1.2,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,32,power_law_1.2,0.04962559938430786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,32,power_law_1.2,0.041407999396324155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,32,power_law_1.2,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,32,power_law_1.2,0.04320639967918396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,32,power_law_1.2,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,32,power_law_1.2,0.05924479961395264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,32,power_law_1.2,0.043532800674438474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,32,power_law_1.2,0.04377599954605103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,32,power_law_1.2,0.06897280216217042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,32,power_law_1.2,0.04894079864025116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,32,power_law_1.2,0.07757440209388733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,32,power_law_1.2,0.049670401215553286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,32,power_law_1.2,0.0990015983581543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,32,power_law_1.2,0.05134080052375793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,32,power_law_1.2,0.12394239902496337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,32,power_law_1.2,0.058847999572753905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,32,power_law_1.2,0.15603840351104736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,32,power_law_1.2,0.060684800148010254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,32,power_law_1.2,0.17900799512863158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,32,power_law_1.2,0.07257599830627441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,32,power_law_1.2,0.268505597114563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,32,power_law_1.2,0.08663679957389832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,32,power_law_1.2,0.4663104057312012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,32,power_law_1.2,0.12781440019607543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,32,power_law_1.2,0.5239424228668212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,32,power_law_1.2,0.1563199996948242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,32,power_law_1.2,0.8097984313964843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,32,power_law_1.2,0.21797120571136475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,32,power_law_1.2,1.9987968444824218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,32,power_law_1.2,0.26369919776916506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,32,power_law_1.2,0.4940608024597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,32,power_law_1.2,0.5196735858917236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,32,power_law_1.2,0.7890687942504883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,32,power_law_1.2,1.293337631225586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,32,power_law_1.2,2.358131217956543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,16,balanced,0.036144000788529716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,16,balanced,0.03836799909671148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,16,balanced,0.03844266633192698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,16,balanced,0.040991999208927155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,16,balanced,0.04475200176239014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,16,balanced,0.044639999667803444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,16,balanced,0.046495998899141945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,16,balanced,0.04664533336957296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,16,balanced,0.044581333796183266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,16,balanced,0.034272000193595886
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,16,balanced,0.04641599953174591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,16,balanced,0.03623999903599421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,16,balanced,0.047135998805363975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,16,balanced,0.0369759996732076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,16,balanced,0.0466186652580897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,16,balanced,0.039173332353432976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,16,balanced,0.04821866750717163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,16,balanced,0.043882668018341064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,16,balanced,0.04483200112978617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,16,balanced,0.04814933240413666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,16,balanced,0.04346133271853129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,16,balanced,0.0506933331489563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,16,balanced,0.04535466432571411
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,16,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,16,balanced,0.04516266783078512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,16,balanced,0.05258666475613912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,16,balanced,0.04529599845409393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,16,balanced,0.04806933303674062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,16,balanced,0.05452266832192739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,16,balanced,0.046613335609436035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,16,balanced,0.049253334601720176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,16,balanced,0.05470400055249532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,16,balanced,0.04888000090916952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,16,balanced,0.06078400214513143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,16,balanced,0.05067733426888784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,16,balanced,0.05195199946562449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,16,balanced,0.06506133576234181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,16,balanced,0.05250666538874308
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,16,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,16,balanced,0.07522666454315186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,16,balanced,0.05858666698137919
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,16,balanced,0.06656000018119812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,16,balanced,0.08702400326728821
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,16,balanced,0.0730560024579366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,16,balanced,0.09532800316810608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,16,balanced,0.10958400368690491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,16,balanced,0.12012267112731934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,16,balanced,0.13029332955678305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,16,balanced,0.16264533003171286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,16,balanced,0.16528532902399698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,16,balanced,0.19989866018295288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,16,power_law_1.01,0.04639999866485596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,16,power_law_1.01,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,16,power_law_1.01,0.039001598954200745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,16,power_law_1.01,0.04267520010471344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,16,power_law_1.01,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,16,power_law_1.01,0.045664000511169436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,16,balanced,0.27212266127268475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,16,power_law_1.01,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,16,power_law_1.01,0.04547840058803558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,16,balanced,0.20578134059906006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,16,power_law_1.01,0.04640640020370483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,16,power_law_1.01,0.04629760086536407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,16,power_law_1.01,0.04619520008563995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,16,power_law_1.01,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,16,power_law_1.01,0.04700160026550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,16,power_law_1.01,0.037299200892448425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,16,power_law_1.01,0.04750719964504242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,16,power_law_1.01,0.03500800132751465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,16,power_law_1.01,0.04840959906578064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,16,power_law_1.01,0.03883520066738129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,16,power_law_1.01,0.051692801713943484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,16,power_law_1.01,0.0400191992521286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,16,balanced,0.3460640112559001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,16,power_law_1.01,0.051820802688598636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,16,power_law_1.01,0.043680000305175784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,16,power_law_1.01,0.05320960283279419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,16,power_law_1.01,0.043136000633239746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,16,power_law_1.01,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,16,power_law_1.01,0.04405759871006012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,16,power_law_1.01,0.043910399079322815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,16,power_law_1.01,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,16,balanced,0.2534880042076111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,16,power_law_1.01,0.04441600143909454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,16,power_law_1.01,0.07471359968185425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,16,power_law_1.01,0.045849600434303285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,16,power_law_1.01,0.08094080090522766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,16,power_law_1.01,0.045817598700523376
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,16,power_law_1.01,0.10021760463714599
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,16,power_law_1.01,0.046105599403381346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,16,power_law_1.01,0.1102463960647583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,16,power_law_1.01,0.047193598747253415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,16,power_law_1.01,0.15079679489135742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,16,power_law_1.01,0.05123839974403381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,16,power_law_1.01,0.17879680395126343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,16,power_law_1.01,0.05281919836997986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,16,power_law_1.01,0.26254079341888426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,16,balanced,0.42132266362508136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,16,power_law_1.01,0.053887999057769774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,16,power_law_1.01,0.3572671890258789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,16,power_law_1.01,0.05987840294837952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,16,power_law_1.01,0.44474239349365235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,16,power_law_1.01,0.06416000127792358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,16,power_law_1.01,0.7257599830627441
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,16,power_law_1.01,0.0769919991493225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,16,power_law_1.01,0.09042559862136841
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,16,power_law_1.01,1.312953567504883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,16,power_law_1.01,0.12975360155105592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,16,power_law_1.01,0.14846080541610718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,16,power_law_1.01,0.20663039684295653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,16,power_law_1.01,0.25626239776611326
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,16,balanced,0.36955734093983966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,16,power_law_1.01,0.39112319946289065
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,16,power_law_1.01,0.4800447940826416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,16,power_law_1.01,0.6306111812591553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,16,power_law_1.01,0.8997695922851563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,16,power_law_1.01,1.833024024963379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,16,balanced,0.6420319875081381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,16,balanced,0.7045919895172119
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,16,balanced,1.2485333283742268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,16,power_law_1.2,0.04547840058803558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,16,power_law_1.2,0.03814400136470795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,16,power_law_1.2,0.039443200826644896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,16,power_law_1.2,0.0411327987909317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,16,power_law_1.2,0.04298880100250244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,16,power_law_1.2,0.042745599150657655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,16,power_law_1.2,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,16,power_law_1.2,0.0367935985326767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,16,power_law_1.2,0.045798400044441225
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,16,power_law_1.2,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,16,power_law_1.2,0.04580479860305786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,16,power_law_1.2,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,16,power_law_1.2,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,16,power_law_1.2,0.03993600010871887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,16,power_law_1.2,0.04713599979877472
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,16,power_law_1.2,0.04258559942245484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,16,power_law_1.2,0.04657920002937317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,16,power_law_1.2,0.043609601259231565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,16,power_law_1.2,0.04702079892158508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,16,power_law_1.2,0.044064000248909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,16,power_law_1.2,0.04767360091209412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,16,power_law_1.2,0.04437119960784912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,16,power_law_1.2,0.04872319996356964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,16,power_law_1.2,0.044537600874900815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,16,power_law_1.2,0.051795202493667605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,16,power_law_1.2,0.044607999920845035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,16,power_law_1.2,0.05236480236053467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,16,power_law_1.2,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,16,power_law_1.2,0.045363199710845944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,16,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,16,power_law_1.2,0.04826239943504333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,16,power_law_1.2,0.05802879929542541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,16,power_law_1.2,0.05146239995956421
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,16,power_law_1.2,0.06607999801635742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,16,power_law_1.2,0.05260800123214722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,16,power_law_1.2,0.07555840015411378
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,16,power_law_1.2,0.05438719987869263
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,16,power_law_1.2,0.0828224003314972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,16,power_law_1.2,0.06101120114326477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,16,power_law_1.2,0.10310399532318115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,16,power_law_1.2,0.06689919829368592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,16,power_law_1.2,0.11132800579071045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,16,power_law_1.2,0.08254719972610473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,16,power_law_1.2,0.18843519687652588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,16,power_law_1.2,0.09848319888114929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,16,power_law_1.2,0.25498878955841064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,8,balanced,0.03589866558710734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,16,power_law_1.2,0.12837120294570922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,16,power_law_1.2,0.32460799217224123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,8,balanced,0.03847466657559077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,16,power_law_1.2,0.16255359649658202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,16,power_law_1.2,0.48995199203491213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,8,balanced,0.0390133336186409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,16,power_law_1.2,0.24481279850006105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,8,balanced,0.044480000933011375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,16,power_law_1.2,0.7363903999328614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,8,balanced,0.048351998130480446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,16,power_law_1.2,0.316921591758728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,16,power_law_1.2,0.8965824127197266
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,8,balanced,0.04753066599369049
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,16,power_law_1.2,0.43656320571899415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,8,balanced,0.049072002371152244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,16,power_law_1.2,2.1963455200195314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,16,power_law_1.2,0.5909311771392822
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,8,balanced,0.05131733417510986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,8,balanced,0.048485333720842995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,16,power_law_1.2,0.9015744209289551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,8,balanced,0.048858667413393654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,16,power_law_1.2,1.3414591789245605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,8,balanced,0.05051200091838837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,8,balanced,0.05082133412361145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,16,power_law_1.2,3.1259328842163088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,8,balanced,0.05212266743183136
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,8,balanced,0.05130666494369507
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,8,balanced,0.054671997825304665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,8,balanced,0.057029331723848976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,8,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,8,balanced,0.05714133381843567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,8,balanced,0.03665599972009659
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,8,balanced,0.03875199953715006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,8,balanced,0.06079466640949249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,8,balanced,0.04177600145339966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,8,balanced,0.04730666677157084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,8,balanced,0.06065600117047628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,8,balanced,0.04742933313051859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,8,balanced,0.04773866633574168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,8,balanced,0.04882133503754934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,8,balanced,0.06955733398596446
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,8,balanced,0.04901866614818573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,8,balanced,0.04910400013128916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,8,balanced,0.049679999550183616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,8,balanced,0.07765866816043854
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,8,balanced,0.04898133377234141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,8,balanced,0.05203733344872793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,8,balanced,0.05073600014050802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,8,balanced,0.096778670946757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,8,balanced,0.05574933191140493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,8,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,8,balanced,0.059024001161257424
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,8,balanced,0.1088053286075592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,8,balanced,0.0634986658891042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,8,balanced,0.06642666459083557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,8,balanced,0.1365173359711965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,8,balanced,0.07809600234031677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,8,balanced,0.08945600191752116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,8,balanced,0.16698666413625082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,8,balanced,0.12201600273450215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,8,balanced,0.14480533202489218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,8,balanced,0.22721600532531738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,8,balanced,0.19643733898798624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,8,balanced,0.24282666047414145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,8,balanced,0.28484266996383667
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,8,balanced,0.3397226730982463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,8,balanced,0.35467199484507245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,8,balanced,0.4377439816792806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,8,balanced,0.5394239823023478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,8,balanced,0.5281013250350952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,8,power_law_1.01,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,8,power_law_1.01,0.04127359986305237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,8,power_law_1.01,0.03959679901599884
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,8,power_law_1.01,0.0430976003408432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,8,power_law_1.01,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,8,power_law_1.01,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,8,power_law_1.01,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,8,balanced,0.8319839636484782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,8,power_law_1.01,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,8,power_law_1.01,0.036857599020004274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,8,power_law_1.01,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,8,power_law_1.01,0.039980798959732056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,8,power_law_1.01,0.04912000000476837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,8,power_law_1.01,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,8,power_law_1.01,0.04578559994697571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,8,power_law_1.01,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,8,power_law_1.01,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,8,power_law_1.01,0.048614400625228885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,8,power_law_1.01,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,8,power_law_1.01,0.05057280063629151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,8,power_law_1.01,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,8,power_law_1.01,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,8,power_law_1.01,0.04779520034790039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,8,power_law_1.01,0.05126399993896484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,8,power_law_1.01,0.049932798743247984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,8,power_law_1.01,0.051577597856521606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,8,power_law_1.01,0.05055360198020935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,8,power_law_1.01,0.05549439787864685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,8,power_law_1.01,0.05109120011329651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,8,power_law_1.01,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,8,power_law_1.01,0.05137280225753784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,8,power_law_1.01,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,8,power_law_1.01,0.05560320019721985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,8,power_law_1.01,0.06272640228271484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,8,power_law_1.01,0.056953597068786624
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,8,power_law_1.01,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,8,power_law_1.01,0.06108160018920898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,8,power_law_1.01,0.08230400085449219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,8,power_law_1.01,0.06807039976119995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,8,power_law_1.01,0.0908352017402649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,8,power_law_1.01,0.07190399765968322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,8,power_law_1.01,0.11348479986190796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,8,power_law_1.01,0.09215360283851623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,8,power_law_1.01,0.12640639543533325
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,8,balanced,1.0019466876983643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,8,power_law_1.01,0.11331199407577515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,8,power_law_1.01,0.1834879994392395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,8,power_law_1.01,0.14544639587402344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,8,power_law_1.01,0.22627201080322265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,8,power_law_1.01,0.1814784049987793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,8,power_law_1.01,0.30599679946899416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,8,power_law_1.01,0.246124792098999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,8,power_law_1.01,0.4544703960418701
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,8,power_law_1.01,0.2840768098831177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,8,power_law_1.01,0.5799359798431396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,8,power_law_1.01,0.4290815830230713
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,8,power_law_1.01,0.7482880115509033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,8,power_law_1.01,0.5643519878387451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,8,power_law_1.01,1.6252927780151367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,8,power_law_1.01,0.71976318359375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,8,power_law_1.01,1.16627197265625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,8,power_law_1.01,2.294265556335449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,8,balanced,1.6306986808776855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,8,power_law_1.2,0.044512000679969785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,8,power_law_1.2,0.038380798697471616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,8,power_law_1.2,0.0390720009803772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,8,power_law_1.2,0.041555199027061465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,8,power_law_1.2,0.04428800046443939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,8,power_law_1.2,0.04665600061416626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,8,power_law_1.2,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,8,power_law_1.2,0.04892799854278564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,8,power_law_1.2,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,8,power_law_1.2,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,8,power_law_1.2,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,8,power_law_1.2,0.05103359818458557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,32,balanced,0.05420800050099691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,8,power_law_1.2,0.0507968008518219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,32,balanced,0.05496533215045929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,8,power_law_1.2,0.05243520140647888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,8,power_law_1.2,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,32,balanced,0.055125330885251365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,8,power_law_1.2,0.05678719878196716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,8,power_law_1.2,0.038022398948669434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,32,balanced,0.0610346645116806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,8,power_law_1.2,0.057792001962661745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,32,balanced,0.06377600133419037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,8,power_law_1.2,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,32,balanced,0.06967466572920482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,8,power_law_1.2,0.05911039710044861
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,32,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,8,power_law_1.2,0.041484799981117246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,32,balanced,0.06966933111349742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,32,balanced,0.05061866839726766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,8,power_law_1.2,0.06654080152511596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,8,power_law_1.2,0.04246399998664856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,32,balanced,0.0514933317899704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,32,balanced,0.06915733218193054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,8,power_law_1.2,0.07468799948692321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,8,power_law_1.2,0.047942399978637695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,32,balanced,0.05764799813429514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,32,balanced,0.06585599978764851
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,8,power_law_1.2,0.0868224024772644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,8,power_law_1.2,0.04744960069656372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,32,balanced,0.05738666653633118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,32,balanced,0.0681333343187968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,32,balanced,0.06131733457247416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,8,power_law_1.2,0.0473471999168396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,8,power_law_1.2,0.09692800045013428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,32,balanced,0.06764799853165944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,32,balanced,0.061349332332611084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,8,power_law_1.2,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,8,power_law_1.2,0.11608959436416626
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,32,balanced,0.06791466474533081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,32,balanced,0.06278933087984721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,8,power_law_1.2,0.05004159808158874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,8,power_law_1.2,0.15887999534606934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,32,balanced,0.06944533189137776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,32,balanced,0.06324266890684764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,8,power_law_1.2,0.049721598625183105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,8,power_law_1.2,0.20780799388885499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,32,balanced,0.06298666695753734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,32,balanced,0.07035199801127116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,8,power_law_1.2,0.05072640180587769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,32,balanced,0.06321600079536438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,8,power_law_1.2,0.2709120035171509
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,32,balanced,0.07313600182533264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,8,power_law_1.2,0.05144320130348205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,32,balanced,0.06548266609509786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,8,power_law_1.2,0.38559999465942385
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,32,balanced,0.07443200051784515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,8,power_law_1.2,0.05278080105781555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,32,balanced,0.0675786683956782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,8,power_law_1.2,0.5360576152801514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,32,balanced,0.07603199779987335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,8,power_law_1.2,0.05711359977722168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,32,balanced,0.06550933420658112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,8,power_law_1.2,0.7440192222595214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,8,power_law_1.2,0.060787200927734375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,32,balanced,0.07106666763623555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,32,balanced,0.08386666576067607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,8,power_law_1.2,0.06206079721450806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,8,power_law_1.2,1.0514495849609375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,32,balanced,0.07108800113201141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,32,balanced,0.08634133140246074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,8,power_law_1.2,0.07059199810028076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,32,balanced,0.07268799841403961
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,8,power_law_1.2,2.1396480560302735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,8,power_law_1.2,0.07409279942512512
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,32,balanced,0.07974933087825775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,32,balanced,0.09845333298047383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,8,power_law_1.2,0.102566397190094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,8,power_law_1.2,0.11424640417099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,32,balanced,0.08479467034339905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,8,power_law_1.2,0.15621119737625122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,32,balanced,0.1067519982655843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,8,power_law_1.2,0.1788032054901123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,32,balanced,0.09999466935793559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,8,power_law_1.2,0.2819648027420044
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,32,balanced,0.12801067034403482
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,32,balanced,0.11609066526095073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,8,power_law_1.2,0.3521023988723755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,8,power_law_1.2,0.5569727897644043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,8,power_law_1.2,0.6552576065063477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,32,balanced,0.16142933567365012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,32,balanced,0.14470932881037393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,8,power_law_1.2,0.8381312370300293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,8,power_law_1.2,1.3768768310546875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,32,balanced,0.18857600291570029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,8,power_law_1.2,2.887788772583008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,32,balanced,0.18047465880711874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,32,balanced,0.2510880033175151
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,32,balanced,0.21278399229049683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,32,balanced,0.309279998143514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,32,balanced,0.27990933259328205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,32,balanced,0.4248480002085368
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,32,balanced,0.3637973467508952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,32,balanced,0.5522773265838623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,32,balanced,0.42583465576171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,32,balanced,0.6650293270746866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,32,balanced,0.6468053261439005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,32,balanced,1.0243199666341145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,32,power_law_1.01,0.06098560094833374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,32,power_law_1.01,0.06303359866142273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,32,power_law_1.01,0.0554751992225647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,32,power_law_1.01,0.058956801891326904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,32,power_law_1.01,0.06119040250778198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,32,power_law_1.01,0.062438398599624634
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,32,power_law_1.01,0.06440960168838501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,32,power_law_1.01,0.06558719873428345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,32,power_law_1.01,0.05798400044441223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,32,power_law_1.01,0.06544640064239501
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,32,power_law_1.01,0.05512319803237915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,32,power_law_1.01,0.06455039978027344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,32,power_law_1.01,0.05240960121154785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,32,power_law_1.01,0.0664896011352539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,32,power_law_1.01,0.05502079725265503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,32,power_law_1.01,0.06615679860115051
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,32,power_law_1.01,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,32,power_law_1.01,0.06682239770889283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,32,power_law_1.01,0.05763840079307556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,32,power_law_1.01,0.06758400201797485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,32,power_law_1.01,0.0604095995426178
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,32,power_law_1.01,0.07166720032691956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,32,power_law_1.01,0.06063359975814819
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,32,power_law_1.01,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,32,power_law_1.01,0.07325440049171447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,32,power_law_1.01,0.06106240153312683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,32,power_law_1.01,0.08321920037269592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,32,power_law_1.01,0.06250240206718445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,32,power_law_1.01,0.08815360069274902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,32,balanced,2.017397403717041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,32,power_law_1.01,0.06291840076446534
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,32,power_law_1.01,0.09654399752616882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,32,balanced,1.2400853633880615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,32,power_law_1.01,0.0633791983127594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,32,power_law_1.01,0.10997120141983033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,32,power_law_1.01,0.06622719764709473
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,32,power_law_1.01,0.12538880109786987
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,32,power_law_1.01,0.07064319849014282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,32,power_law_1.01,0.1524608016014099
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,32,power_law_1.01,0.07177600264549255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,32,power_law_1.01,0.18691200017929077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,32,power_law_1.01,0.07461119890213012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,32,power_law_1.01,0.24352641105651857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,32,power_law_1.01,0.08561279773712158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,32,power_law_1.01,0.3079551935195923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,32,power_law_1.01,0.09536640048027038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,32,power_law_1.01,0.4382463932037354
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,32,power_law_1.01,0.12030719518661499
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,32,power_law_1.01,0.512390422821045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,32,power_law_1.01,0.1415168046951294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,32,power_law_1.01,0.6008639812469483
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,32,power_law_1.01,0.17568639516830445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,32,power_law_1.01,0.9513279914855957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,32,power_law_1.01,0.21084160804748536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,32,power_law_1.01,0.29185919761657714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,32,power_law_1.01,1.9651391983032227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,32,power_law_1.01,0.3630784034729004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,32,power_law_1.01,0.5495232105255127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,32,power_law_1.01,0.7450240135192872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,32,power_law_1.01,0.8623616218566894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,32,power_law_1.01,1.3494720458984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,32,power_law_1.01,2.944972801208496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,32,power_law_1.2,0.058937597274780276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,32,power_law_1.2,0.05411199927330017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,32,power_law_1.2,0.050246399641036985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,32,power_law_1.2,0.055404800176620486
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,32,power_law_1.2,0.05720959901809693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,32,power_law_1.2,0.06566399931907654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,32,power_law_1.2,0.05752320289611816
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,32,power_law_1.2,0.056569600105285646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,32,power_law_1.2,0.055232000350952146
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,32,power_law_1.2,0.05899519920349121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,32,power_law_1.2,0.060115200281143186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,32,power_law_1.2,0.05985280275344849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,32,power_law_1.2,0.06142079830169678
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,32,power_law_1.2,0.06120319962501526
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,32,power_law_1.2,0.06220160126686096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,32,power_law_1.2,0.06327040195465088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,32,power_law_1.2,0.06344320178031922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,32,power_law_1.2,0.06607360243797303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,32,power_law_1.2,0.06309120059013366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,32,power_law_1.2,0.06572800278663635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,32,power_law_1.2,0.06593919992446899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,32,power_law_1.2,0.06789759993553161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,32,power_law_1.2,0.06580479741096497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,32,power_law_1.2,0.0659775972366333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,16,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,32,power_law_1.2,0.06959360241889953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,32,power_law_1.2,0.06686080098152161
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,16,balanced,0.05378133555253347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,32,power_law_1.2,0.07078400254249573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,32,power_law_1.2,0.06646400094032287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,16,balanced,0.05514133473237356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,16,balanced,0.06284800171852112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,32,power_law_1.2,0.07645440101623535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,32,power_law_1.2,0.06764159798622131
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,16,balanced,0.06403733293215434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,32,power_law_1.2,0.08894720077514648
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,32,power_law_1.2,0.06895359754562377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,16,balanced,0.0694400022427241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,32,power_law_1.2,0.10012160539627075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,32,power_law_1.2,0.07793920040130616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,16,balanced,0.0694400022427241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,32,power_law_1.2,0.12303359508514404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,32,power_law_1.2,0.08051199913024902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,16,balanced,0.06987200180689494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,32,power_law_1.2,0.15340800285339357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,32,power_law_1.2,0.08272640109062195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,16,balanced,0.06681600213050842
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,16,balanced,0.07035199801127116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,32,power_law_1.2,0.18917759656906127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,32,power_law_1.2,0.0866815984249115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,16,balanced,0.06883733471234639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,32,power_law_1.2,0.23978240489959718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,32,power_law_1.2,0.09989759922027588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,16,balanced,0.06938133140405019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,32,power_law_1.2,0.11723519563674926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,32,power_law_1.2,0.3247040033340454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,16,balanced,0.0709386666615804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,32,power_law_1.2,0.13012479543685912
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,32,power_law_1.2,0.4369919776916504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,16,balanced,0.07169599831104279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,32,power_law_1.2,0.16556799411773682
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,32,power_law_1.2,0.6258624076843262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,16,balanced,0.07521600027879079
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,32,power_law_1.2,0.1968191981315613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,16,balanced,0.048309331138928734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,32,power_law_1.2,0.8237312316894532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,16,balanced,0.07791466514269511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,16,balanced,0.04933333396911621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,32,power_law_1.2,0.2624128103256226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,32,power_law_1.2,1.0479935646057128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,16,balanced,0.05123733480771383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,16,balanced,0.0792799989382426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,32,power_law_1.2,0.3221951961517334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,16,balanced,0.05726933479309082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,32,power_law_1.2,1.72936954498291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,32,power_law_1.2,0.47772798538208006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,16,balanced,0.059104000528653465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,16,balanced,0.08530666430791219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,16,balanced,0.06409066418806712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,32,power_law_1.2,0.690393590927124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,32,power_law_1.2,3.69617919921875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,16,balanced,0.06576533118883769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,16,balanced,0.09300800164540608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,32,power_law_1.2,0.8198783874511719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,16,balanced,0.0658079981803894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,16,balanced,0.06367999811967213
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,32,power_law_1.2,1.3799424171447754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,16,balanced,0.1037493348121643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,16,balanced,0.06542400022347768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,16,balanced,0.06545599798361461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,32,power_law_1.2,3.456051254272461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,16,balanced,0.11596266428629558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,16,balanced,0.0666293352842331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,16,balanced,0.06823466718196869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,16,balanced,0.06738133231798808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,16,balanced,0.07344000041484833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,16,balanced,0.14300800363222757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,16,balanced,0.0740586668252945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,16,balanced,0.07537066439787547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,16,balanced,0.08410666386286418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,16,balanced,0.160181333621343
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,16,balanced,0.09186133742332458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,16,balanced,0.10937066872914632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,16,balanced,0.20272533098856607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,16,balanced,0.12965866923332214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,16,balanced,0.17465599377950033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,16,balanced,0.24104533592859903
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,16,balanced,0.2023893396059672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,16,balanced,0.2708746592203776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,16,balanced,0.32340800762176514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,16,balanced,0.33747732639312744
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,16,balanced,0.4672693411509196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,16,balanced,0.42018131415049237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,16,balanced,0.6064800024032593
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,16,balanced,0.49420801798502606
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,16,balanced,0.7352426846822103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,16,balanced,0.7577599684397379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,16,power_law_1.01,0.06938239932060242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,16,power_law_1.01,0.0636672019958496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,16,power_law_1.01,0.05687680244445801
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,16,power_law_1.01,0.06056960225105286
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,16,balanced,1.1393067042032878
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,16,power_law_1.01,0.0637503981590271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,16,power_law_1.01,0.06727679967880248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,16,power_law_1.01,0.06848000288009644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,16,power_law_1.01,0.0673088014125824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,16,power_law_1.01,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,16,power_law_1.01,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,16,power_law_1.01,0.06963840126991272
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,16,power_law_1.01,0.07088639736175537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,16,power_law_1.01,0.062041598558425906
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,16,power_law_1.01,0.07188479900360108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,16,power_law_1.01,0.07139840126037597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,16,power_law_1.01,0.05922560095787048
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,16,power_law_1.01,0.08142719864845276
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,16,power_law_1.01,0.05306879878044128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,16,power_law_1.01,0.08223999738693237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,16,power_law_1.01,0.058246397972106935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,16,power_law_1.01,0.08765439987182617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,16,power_law_1.01,0.05799679756164551
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,16,power_law_1.01,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,16,power_law_1.01,0.0624064028263092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,16,power_law_1.01,0.10469119548797608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,16,power_law_1.01,0.06312320232391358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,16,power_law_1.01,0.11779199838638306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,16,power_law_1.01,0.06377599835395813
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,16,power_law_1.01,0.1340288043022156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,16,power_law_1.01,0.06462720036506653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,16,power_law_1.01,0.1580288052558899
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,16,power_law_1.01,0.0645695984363556
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,16,power_law_1.01,0.20212481021881104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,16,power_law_1.01,0.06527360081672669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,16,power_law_1.01,0.2599040031433105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,16,power_law_1.01,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,16,power_law_1.01,0.06641280055046081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,16,power_law_1.01,0.3399296045303345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,16,power_law_1.01,0.06787199974060058
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,16,power_law_1.01,0.495359992980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,16,power_law_1.01,0.0716863989830017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,16,power_law_1.01,0.6188543796539306
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,16,power_law_1.01,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,16,power_law_1.01,0.7080128192901611
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,16,power_law_1.01,0.0789247989654541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,16,power_law_1.01,1.1687935829162597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,16,power_law_1.01,0.089683198928833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,16,power_law_1.01,2.369510459899902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,16,power_law_1.01,0.10281599760055542
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,16,power_law_1.01,0.1314303994178772
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,16,power_law_1.01,0.14941439628601075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,16,power_law_1.01,0.19253120422363282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,16,power_law_1.01,0.23676800727844238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,16,power_law_1.01,0.3146944046020508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,16,balanced,1.4845280647277832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,16,power_law_1.01,0.40497279167175293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,16,balanced,2.256506601969401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,16,power_law_1.01,0.6003967761993408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,16,power_law_1.01,0.7605631828308106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,16,power_law_1.01,1.0035327911376952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,16,power_law_1.01,1.4663423538208007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,16,power_law_1.01,3.094432067871094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,16,power_law_1.2,0.06972799897193908
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,16,power_law_1.2,0.05813120007514953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,16,power_law_1.2,0.05624319911003113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,16,power_law_1.2,0.059385597705841064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,16,power_law_1.2,0.06382719874382019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,16,power_law_1.2,0.06551039814949036
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,16,power_law_1.2,0.06727679967880248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,16,power_law_1.2,0.060115200281143186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,16,power_law_1.2,0.06880000233650208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,16,power_law_1.2,0.05203199982643127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,16,power_law_1.2,0.06869760155677795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,16,power_law_1.2,0.05161600112915039
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,16,power_law_1.2,0.06876800060272217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,16,power_law_1.2,0.0574783980846405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,8,balanced,0.052576000491778054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,16,power_law_1.2,0.07103360295295716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,16,power_law_1.2,0.05867519974708557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,8,balanced,0.05499200026194254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,16,power_law_1.2,0.07325440049171447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,16,power_law_1.2,0.061427199840545656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,8,balanced,0.057248001297314964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,16,power_law_1.2,0.0706816017627716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,8,balanced,0.06297066807746887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,16,power_law_1.2,0.06357120275497437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,8,balanced,0.07186666627724965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,16,power_law_1.2,0.07288960218429566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,16,power_law_1.2,0.06498559713363647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,8,balanced,0.0783786674340566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,16,power_law_1.2,0.07599999904632568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,16,power_law_1.2,0.06497920155525208
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,8,balanced,0.07792533437410991
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,16,power_law_1.2,0.08447999954223633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,16,power_law_1.2,0.06455039978027344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,8,balanced,0.07825066645940144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,16,power_law_1.2,0.08695039749145508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,8,balanced,0.04826666911443075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,16,power_law_1.2,0.06590080261230469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,8,balanced,0.07569600145022075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,8,balanced,0.05120533208052317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,16,power_law_1.2,0.10117759704589843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,16,power_law_1.2,0.06662399768829345
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,8,balanced,0.0759680022795995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,8,balanced,0.053770666321118675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,16,power_law_1.2,0.10657279491424561
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,16,power_law_1.2,0.06538879871368408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,8,balanced,0.06115733087062836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,8,balanced,0.07763200004895528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,16,power_law_1.2,0.12655999660491943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,8,balanced,0.06341333190600078
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,16,power_law_1.2,0.06730239987373351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,8,balanced,0.07770666480064392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,8,balanced,0.06994133194287618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,16,power_law_1.2,0.1378432035446167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,16,power_law_1.2,0.07420160174369812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,8,balanced,0.07638933261235555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,8,balanced,0.0711413323879242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,16,power_law_1.2,0.18486399650573732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,16,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,8,balanced,0.07891199986139934
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,8,balanced,0.07025599976380666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,16,power_law_1.2,0.22498559951782227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,16,power_law_1.2,0.08013439774513245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,8,balanced,0.07039999961853027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,8,balanced,0.08222400148709615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,16,power_law_1.2,0.09336959719657897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,16,power_law_1.2,0.30572800636291503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,8,balanced,0.07214400172233582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,8,balanced,0.08568533261617024
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,16,power_law_1.2,0.10355199575424194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,16,power_law_1.2,0.3892352104187012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,8,balanced,0.07147733370463054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,8,balanced,0.07201600074768066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,8,balanced,0.0902346670627594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,16,power_law_1.2,0.1332800030708313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,16,power_law_1.2,0.48285441398620604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,8,balanced,0.07401066521803538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,16,power_law_1.2,0.15194239616394042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,16,power_law_1.2,0.8190400123596191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,8,balanced,0.09903466701507568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,8,balanced,0.07372266550858815
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,16,power_law_1.2,0.1956287980079651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,16,power_law_1.2,0.9342592239379883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,8,balanced,0.08080533146858215
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,8,balanced,0.10593600074450175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,16,power_law_1.2,0.25284481048583984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,16,power_law_1.2,1.565171241760254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,8,balanced,0.08156266808509827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,16,power_law_1.2,0.356876802444458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,8,balanced,0.11959466338157654
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,16,power_law_1.2,0.47330560684204104
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,8,balanced,0.08372799555460612
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,16,power_law_1.2,3.176838493347168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,16,power_law_1.2,0.7341887950897217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,8,balanced,0.09567999839782715
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,8,balanced,0.13081600268681845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,16,power_law_1.2,0.8980992317199707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,8,balanced,0.10187199711799622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,16,power_law_1.2,1.0773759841918946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,16,power_law_1.2,1.6908735275268554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,8,balanced,0.12924266854921976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,8,balanced,0.16460266709327698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,16,power_law_1.2,3.853977584838867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,8,balanced,0.14896000425020853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,8,balanced,0.18683733542760214
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,8,balanced,0.19740800062815347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,8,balanced,0.24654932816823324
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,8,balanced,0.2323466738065084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,8,balanced,0.31380265951156616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,8,balanced,0.3163520097732544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,8,balanced,0.40564266840616864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,8,balanced,0.4185813268025716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,8,balanced,0.5641760031382242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,8,balanced,0.5352960030237833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,8,balanced,0.7255840301513672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,8,balanced,0.6450453201929728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,8,balanced,0.8916373252868652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,8,balanced,0.9922880331675211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,8,power_law_1.01,0.06715520024299622
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,8,power_law_1.01,0.06679040193557739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,8,power_law_1.01,0.05927680134773254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,8,balanced,1.3833813667297363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,8,power_law_1.01,0.06446719765663148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,8,power_law_1.01,0.06682239770889283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,8,power_law_1.01,0.07139840126037597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,8,power_law_1.01,0.07473919987678528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,8,power_law_1.01,0.0609279990196228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,8,power_law_1.01,0.07640320062637329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,8,power_law_1.01,0.06219519972801209
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,8,power_law_1.01,0.07575680017471313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,8,power_law_1.01,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,8,power_law_1.01,0.07617920041084289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,8,power_law_1.01,0.05717759728431702
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,8,power_law_1.01,0.07762560248374939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,8,power_law_1.01,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,8,power_law_1.01,0.08090239763259888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,8,power_law_1.01,0.06690559983253479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,8,power_law_1.01,0.07693439722061157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,8,power_law_1.01,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,8,power_law_1.01,0.08102399706840516
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,8,power_law_1.01,0.06995199918746949
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,8,power_law_1.01,0.08855040073394775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,8,power_law_1.01,0.07124480009078979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,8,power_law_1.01,0.09190400242805481
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,8,power_law_1.01,0.07155839800834655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,8,power_law_1.01,0.09501439929008484
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,8,power_law_1.01,0.07254400253295898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,8,power_law_1.01,0.10431360006332398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,8,power_law_1.01,0.07216640114784241
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,8,power_law_1.01,0.1162559986114502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,8,power_law_1.01,0.07444480061531067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,8,power_law_1.01,0.1355839967727661
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,8,power_law_1.01,0.07672960162162781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,8,power_law_1.01,0.15031039714813232
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,8,power_law_1.01,0.08061439990997314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,8,power_law_1.01,0.18154239654541016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,8,power_law_1.01,0.0821183979511261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,8,power_law_1.01,0.22028160095214844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,8,power_law_1.01,0.08752639889717102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,8,balanced,1.9503092765808105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,8,power_law_1.01,0.10382720232009887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,8,power_law_1.01,0.30655360221862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,8,power_law_1.01,0.11949440240859985
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,8,power_law_1.01,0.4001728057861328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,8,power_law_1.01,0.1429952025413513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,8,power_law_1.01,0.5237184047698975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,8,power_law_1.01,0.17080960273742676
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,8,power_law_1.01,0.7402304172515869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,8,power_law_1.01,0.21553280353546142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,8,power_law_1.01,0.9072447776794433
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,8,power_law_1.01,0.28606719970703126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,8,power_law_1.01,1.2916095733642579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,8,power_law_1.01,0.3817087888717651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,8,power_law_1.01,2.9002048492431642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,8,power_law_1.01,0.47342720031738283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,8,power_law_1.01,0.6717184066772461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,8,power_law_1.01,0.9388928413391113
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,8,power_law_1.01,1.1237312316894532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,8,power_law_1.01,1.7762432098388672
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,8,power_law_1.01,3.50052490234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,8,balanced,2.7332639694213867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,8,power_law_1.2,0.06616320013999939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,8,power_law_1.2,0.06056320071220398
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,8,power_law_1.2,0.05913599729537964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,8,power_law_1.2,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,8,power_law_1.2,0.06058239936828613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,8,power_law_1.2,0.05356799960136414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,8,power_law_1.2,0.06357759833335877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,8,power_law_1.2,0.0573311984539032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,8,power_law_1.2,0.0655232012271881
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,128,balanced,0.06263466676076253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,8,power_law_1.2,0.05999360084533691
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,8,power_law_1.2,0.07403519749641418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,128,balanced,0.06743466854095459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,8,power_law_1.2,0.06754559874534607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,8,power_law_1.2,0.07534719705581665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,128,balanced,0.06563733518123627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,128,balanced,0.06904000043869019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,8,power_law_1.2,0.06847360134124755
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,8,power_law_1.2,0.07612800002098083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,128,balanced,0.0668639987707138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,8,power_law_1.2,0.06931840181350708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,8,power_law_1.2,0.07827200293540955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,128,balanced,0.07813866436481476
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,8,power_law_1.2,0.07021440267562866
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,8,power_law_1.2,0.08018559813499451
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,128,balanced,0.08947733044624329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,8,power_law_1.2,0.07188479900360108
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,128,balanced,0.08987733721733093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,8,power_law_1.2,0.07752320170402527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,128,balanced,0.08832533160845439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,8,power_law_1.2,0.07311360239982605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,8,power_law_1.2,0.07809280157089234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,128,balanced,0.09084799885749817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,8,power_law_1.2,0.07388160228729249
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,8,power_law_1.2,0.07969279885292054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,128,balanced,0.0944106678167979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,8,power_law_1.2,0.07508479952812194
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,8,power_law_1.2,0.08446720242500305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,128,balanced,0.10139733552932739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,128,balanced,0.05637866755326589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,8,power_law_1.2,0.07734400033950806
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,8,power_law_1.2,0.08810880184173583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,128,balanced,0.05718400080998739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,128,balanced,0.09618133306503296
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,8,power_law_1.2,0.08130559921264649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,8,power_law_1.2,0.09303680062294006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,128,balanced,0.05750399827957153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,128,balanced,0.09579199552536011
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,128,balanced,0.06116800010204315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,8,power_law_1.2,0.08423039913177491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,8,power_law_1.2,0.09934080243110657
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,128,balanced,0.099263995885849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,128,balanced,0.057631999254226685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,8,power_law_1.2,0.0923583984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,8,power_law_1.2,0.10837759971618652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,128,balanced,0.10632000366846721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,128,balanced,0.06668800115585327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,8,power_law_1.2,0.10968960523605346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,8,power_law_1.2,0.11661440134048462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,128,balanced,0.07190933326880138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,128,balanced,0.10839466253916423
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,8,power_law_1.2,0.12375680208206177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,128,balanced,0.07121066749095917
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,8,power_law_1.2,0.1409600019454956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,128,balanced,0.07317866881688435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,128,balanced,0.1116480032602946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,8,power_law_1.2,0.15352959632873536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,8,power_law_1.2,0.16360960006713868
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,128,balanced,0.07006399830182393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,8,power_law_1.2,0.17566720247268677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,8,power_law_1.2,0.1975167989730835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,128,balanced,0.1195146640141805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,128,balanced,0.07307733098665874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,8,power_law_1.2,0.2327552080154419
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,8,power_law_1.2,0.27250559329986573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,128,balanced,0.07383466760317485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,8,power_law_1.2,0.32376320362091066
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,128,balanced,0.13011733690897623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,8,power_law_1.2,0.33959040641784666
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,128,balanced,0.07423999905586243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,8,power_law_1.2,0.4094848155975342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,8,power_law_1.2,0.4262976169586182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,128,balanced,0.07391466697057088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,8,power_law_1.2,0.520691204071045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,128,balanced,0.08063466846942902
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,128,balanced,0.14273599783579508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,8,power_law_1.2,0.6403391838073731
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,8,power_law_1.2,0.7349247932434082
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,128,balanced,0.08038933575153351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,8,power_law_1.2,0.8834624290466309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,8,power_law_1.2,0.9821375846862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,8,power_law_1.2,1.0199616432189942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,128,balanced,0.17030400037765503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,8,power_law_1.2,1.1654784202575683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,8,power_law_1.2,1.6736000061035157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,8,power_law_1.2,2.1375999450683594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,8,power_law_1.2,3.8753662109375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,8,power_law_1.2,3.954790496826172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,128,balanced,0.194106658299764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,128,balanced,0.24713599681854248
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,128,balanced,0.30082666873931885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,128,balanced,0.402672012646993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,128,balanced,0.5076160033543905
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,128,balanced,0.6316106716791788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,128,balanced,0.9472906589508057
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,128,balanced,0.05445333321889242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,128,power_law_1.01,0.08236799836158752
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,128,balanced,0.05548266569773356
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,128,balanced,0.05728533367315928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,128,power_law_1.01,0.08275840282440186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,128,balanced,0.061066667238871254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,128,power_law_1.01,0.08108159899711609
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,128,balanced,0.05778133372465769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,128,balanced,0.06806933383146922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,128,power_law_1.01,0.07639679908752442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,128,balanced,0.07807999849319458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,128,balanced,0.07315200070540111
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,128,power_law_1.01,0.08334720134735107
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,128,balanced,0.0720000018676122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,128,power_law_1.01,0.08515200018882751
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,128,balanced,0.06994133194287618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,128,balanced,0.07353599866231282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,128,power_law_1.01,0.08643839955329895
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,128,balanced,0.07576533158620198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,128,balanced,1.8357812563578289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,128,power_law_1.01,0.0905344009399414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,128,balanced,0.07150400181611379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,128,balanced,0.07397333284219106
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,128,power_law_1.01,0.09147520065307617
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,128,balanced,0.08986666798591614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,128,power_law_1.01,0.09557120203971863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,128,balanced,0.08183466891447704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,128,power_law_1.01,0.09093120098114013
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,128,balanced,0.09154666463534038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,128,power_law_1.01,0.09470720291137695
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,128,balanced,0.09612799684206645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,128,power_law_1.01,0.0923583984375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,128,balanced,0.08956266442934673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,128,balanced,0.10622400045394897
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,128,power_law_1.01,0.09324799776077271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,128,balanced,0.10041600465774536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,128,power_law_1.01,0.0996288001537323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,128,balanced,0.11753066380818684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,128,balanced,0.10923199852307637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,128,power_law_1.01,0.09639040231704712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,128,balanced,0.13895466923713684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,128,balanced,0.1350879967212677
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,128,power_law_1.01,0.10361599922180176
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,128,power_law_1.01,0.11026560068130493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,128,balanced,0.15523733695348105
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,128,power_law_1.01,0.11782399415969849
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,128,balanced,0.16662933429082236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,128,power_law_1.01,0.1334272027015686
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,128,balanced,0.19806933403015137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,128,power_law_1.01,0.1509055972099304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,128,power_law_1.01,0.19160959720611573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,128,balanced,0.1943999926249186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,128,power_law_1.01,0.22890880107879638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,128,balanced,0.23704532782236734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,128,power_law_1.01,0.295251202583313
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,128,power_law_1.01,0.37879679203033445
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,128,balanced,0.2545173366864522
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,128,power_law_1.01,0.5468160152435303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,128,balanced,0.3184906641642253
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,128,power_law_1.01,0.680614423751831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,128,power_law_1.01,0.8449151992797852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,128,power_law_1.01,1.2737664222717284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,128,balanced,0.32171199719111127
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,128,balanced,0.4012586673100789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,128,power_law_1.01,2.6088640213012697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,128,balanced,0.4398239850997925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,128,balanced,0.5607893466949463
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,128,balanced,0.5634239912033081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,128,balanced,0.7223520278930664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,128,balanced,0.8923467000325521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,128,balanced,0.6936853726704916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,128,balanced,1.3751840591430664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,128,balanced,1.0597866376241047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,128,power_law_1.01,0.06474239826202392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,128,power_law_1.01,0.06105599999427795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,128,power_law_1.01,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,128,power_law_1.01,0.06101120114326477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,128,power_law_1.01,0.0634112000465393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,128,power_law_1.01,0.0650111973285675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,128,power_law_1.01,0.06570240259170532
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,128,power_law_1.01,0.06691840291023254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,128,power_law_1.01,0.06925439834594727
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,128,power_law_1.01,0.0700543999671936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,128,power_law_1.01,0.07020800113677979
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,128,power_law_1.01,0.069760000705719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,128,power_law_1.01,0.07036160230636597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,128,power_law_1.01,0.07212160229682922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,128,power_law_1.01,0.08076159954071045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,128,power_law_1.01,0.0794048011302948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,128,power_law_1.01,0.08528000116348267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,128,power_law_1.01,0.09909120202064514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,128,power_law_1.01,0.11202559471130372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,128,power_law_1.01,0.13968640565872192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,128,power_law_1.01,0.0698751986026764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,128,power_law_1.01,0.16246399879455567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,128,power_law_1.01,0.06547200083732604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,128,power_law_1.01,0.21098880767822265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,128,power_law_1.01,0.0654591977596283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,128,power_law_1.01,0.258188796043396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,128,balanced,2.701840082804362
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,128,power_law_1.01,0.06032639741897583
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,128,balanced,2.0685973167419434
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,128,power_law_1.01,0.3560767889022827
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,128,power_law_1.01,0.06297600269317627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,128,power_law_1.01,0.4622528076171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,128,power_law_1.01,0.06459519863128663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,128,power_law_1.01,0.6531712055206299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,128,power_law_1.01,0.06761599779129028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,128,power_law_1.01,0.8289152145385742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,128,power_law_1.01,0.07081599831581116
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,128,power_law_1.01,1.0785728454589845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,128,power_law_1.01,0.07114239931106567
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,128,power_law_1.01,1.624166488647461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,128,power_law_1.01,0.06917759776115417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,128,power_law_1.01,0.07220479846000671
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,128,power_law_1.01,3.3401920318603517
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,128,power_law_1.01,0.07425280213356018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,128,power_law_1.01,0.07193599939346314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,128,power_law_1.01,0.07349759936332703
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,128,power_law_1.01,0.07612800002098083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,128,power_law_1.01,0.08011519908905029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,128,power_law_1.01,0.0849407970905304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,128,power_law_1.01,0.092467200756073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,128,power_law_1.01,0.09673600196838379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,128,power_law_1.01,0.12498559951782226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,128,power_law_1.01,0.14447360038757323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,128,power_law_1.01,0.18053120374679565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,128,power_law_1.01,0.21738879680633544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,128,power_law_1.01,0.2933568000793457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,128,power_law_1.01,0.35942399501800537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,128,power_law_1.01,0.5250495910644531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,128,power_law_1.01,0.6669951915740967
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,128,power_law_1.01,0.8125503540039063
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,128,power_law_1.01,1.2959936141967774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,128,power_law_1.01,2.5469568252563475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,128,power_law_1.2,0.06257920265197754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,128,power_law_1.2,0.08262400031089782
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,128,power_law_1.2,0.06106879711151123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,128,power_law_1.2,0.08342400193214417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,128,power_law_1.2,0.06028159856796265
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,128,power_law_1.2,0.08223999738693237
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,128,power_law_1.2,0.06791679859161377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,128,power_law_1.2,0.06316159963607788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,128,power_law_1.2,0.07656319737434387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,128,power_law_1.2,0.06473600268363952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,128,power_law_1.2,0.0650816023349762
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,128,power_law_1.2,0.08312960267066956
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,128,power_law_1.2,0.06649600267410279
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,128,power_law_1.2,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,128,power_law_1.2,0.08741120100021363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,128,power_law_1.2,0.061459201574325564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,128,power_law_1.2,0.06542720198631287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,128,power_law_1.2,0.08774399757385254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,128,power_law_1.2,0.06581760048866273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,128,power_law_1.2,0.06835200190544129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,128,power_law_1.2,0.08825600147247314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,128,power_law_1.2,0.06421759724617004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,128,power_law_1.2,0.06870399713516236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,128,power_law_1.2,0.0926144003868103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,128,power_law_1.2,0.06955519914627076
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,128,power_law_1.2,0.06436480283737182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,128,power_law_1.2,0.08914560079574585
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,128,power_law_1.2,0.07048320174217224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,128,power_law_1.2,0.06605439782142639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,128,power_law_1.2,0.08885759711265565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,128,power_law_1.2,0.07141119837760926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,128,power_law_1.2,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,128,power_law_1.2,0.08848000168800355
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,128,power_law_1.2,0.06845440268516541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,128,power_law_1.2,0.06818559765815735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,128,power_law_1.2,0.09464960098266602
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,128,power_law_1.2,0.07541120052337646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,128,power_law_1.2,0.06777600049972535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,128,power_law_1.2,0.09281280040740966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,128,power_law_1.2,0.07715839743614197
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,128,power_law_1.2,0.07111039757728577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,128,power_law_1.2,0.0970304012298584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,128,power_law_1.2,0.08165119886398316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,128,power_law_1.2,0.07096959948539734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,128,power_law_1.2,0.09479680061340331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,128,power_law_1.2,0.0859391987323761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,128,power_law_1.2,0.07193599939346314
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,128,power_law_1.2,0.0992255985736847
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,128,power_law_1.2,0.10289280414581299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,128,power_law_1.2,0.1120959997177124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,128,power_law_1.2,0.11632640361785888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,128,power_law_1.2,0.1201856017112732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,128,power_law_1.2,0.14216320514678954
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,128,power_law_1.2,0.13781119585037233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,128,power_law_1.2,0.16832000017166138
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,128,power_law_1.2,0.16368639469146729
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,128,power_law_1.2,0.22167680263519288
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,128,power_law_1.2,0.21191039085388183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,128,power_law_1.2,0.2533695936203003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,128,power_law_1.2,0.2752000093460083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,128,power_law_1.2,0.3282880067825317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,128,power_law_1.2,0.3938431978225708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,128,power_law_1.2,0.4963967800140381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,128,power_law_1.2,0.42444801330566406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,128,power_law_1.2,0.6841023921966553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,128,power_law_1.2,0.6541759967803955
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,128,power_law_1.2,0.8241408348083497
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,128,power_law_1.2,0.9036224365234375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,128,power_law_1.2,1.0058752059936524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,128,power_law_1.2,1.175001621246338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,64,balanced,0.06069866816202799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,128,power_law_1.2,1.6169471740722656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,64,balanced,0.06062399844328562
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,128,power_law_1.2,1.7909568786621093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,64,balanced,0.061477333307266235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,64,balanced,0.06347733239332835
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,128,power_law_1.2,3.81610221862793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,128,power_law_1.2,3.7646976470947267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,64,balanced,0.06758399804433186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,64,balanced,0.08022400240103404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,64,balanced,0.09085333347320557
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,64,balanced,0.09011733531951904
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,64,balanced,0.08718400200208028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,64,balanced,0.08986666798591614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,64,balanced,0.0920799970626831
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,64,balanced,0.09246399998664856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,64,balanced,0.09452266494433086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,64,balanced,0.09391466776529948
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,64,balanced,0.0962559978167216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,64,balanced,0.10108266274134318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,64,balanced,0.10232533017794292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,64,balanced,0.10828800002733867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,64,balanced,0.1169599990049998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,64,balanced,0.12959466377894083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,64,balanced,0.1418400009473165
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,64,balanced,0.1704639991124471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,64,balanced,0.19580266873041788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,64,balanced,0.24921600023905435
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,64,balanced,0.3039626677831014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,64,balanced,0.4093386729558309
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,64,balanced,0.5177226861317953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,64,balanced,0.642186681429545
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,64,balanced,0.9649439652760824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,128,power_law_1.2,0.07645440101623535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,128,power_law_1.2,0.07908480167388916
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,64,balanced,0.054416000843048096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,64,balanced,0.054511999090512596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,128,power_law_1.2,0.0821951985359192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,64,balanced,0.050586665670077004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,64,balanced,0.05464000006516775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,64,balanced,0.05313600103060404
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,128,power_law_1.2,0.09393280148506164
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,64,balanced,0.058837334314982094
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,64,balanced,0.0517493337392807
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,64,balanced,0.059631998340288796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,64,balanced,0.05457599957784017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,128,power_law_1.2,0.10317440032958984
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,64,balanced,0.06758933266003926
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,64,balanced,0.059194669127464294
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,128,power_law_1.2,0.12672640085220338
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,64,balanced,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,64,balanced,0.061573331554730736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,64,balanced,0.07389333347479503
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,64,balanced,0.06679999828338623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,128,power_law_1.2,0.14653439521789552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,64,balanced,0.07132799923419952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,64,balanced,0.06931733091672261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,128,power_law_1.2,0.1940608024597168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,64,balanced,0.07601599891980489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,64,balanced,0.07055999835332234
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,64,balanced,0.06709333260854085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,64,balanced,0.07307733098665874
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,128,power_law_1.2,0.23116800785064698
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,64,balanced,0.07162133355935414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,64,balanced,0.07361066838105519
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,128,power_law_1.2,0.3179136037826538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,64,balanced,0.07156266768773396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,64,balanced,0.07747733096281688
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,64,balanced,0.07114666700363159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,64,balanced,0.07647466659545898
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,128,power_law_1.2,0.4205952167510986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,64,balanced,0.07517333328723907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,64,balanced,0.08756267031033833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,128,power_law_1.2,0.5657663822174073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,64,balanced,0.08094933132330577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,64,balanced,0.08337600032488506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,128,power_law_1.2,0.7455552101135254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,64,balanced,0.08184533317883809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,64,balanced,0.08706133564313252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,128,power_law_1.2,0.916659164428711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,64,balanced,0.09000000357627869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,64,balanced,0.09362133344014485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,64,balanced,0.09523733456929524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,128,power_law_1.2,1.4482111930847168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,64,balanced,0.10437333583831787
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,64,balanced,0.10780800382296245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,128,power_law_1.2,3.1930112838745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,64,balanced,0.11801066994667053
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,64,balanced,0.13552533586819968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,64,balanced,0.13703999916712442
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,64,balanced,0.15770666797955832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,64,balanced,0.16953066984812418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,64,balanced,0.20016000668207803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,64,balanced,1.8784853617350261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,64,balanced,0.1982133388519287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,64,balanced,0.2407840092976888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,64,balanced,0.25999999046325684
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,64,balanced,0.3230453332265218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,64,balanced,0.33286933104197186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,64,balanced,0.40810132026672363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,64,balanced,0.45229331652323407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,64,balanced,0.5727893511454264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,64,balanced,0.5793279806772867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,64,balanced,0.7360906600952148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,64,balanced,0.710640033086141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,64,balanced,0.9097973505655924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,64,balanced,1.085040012995402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,64,balanced,1.3996159235636394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,64,power_law_1.01,0.08652799725532531
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,64,power_law_1.01,0.08395519852638245
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,64,power_law_1.01,0.08355839848518372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,64,power_law_1.01,0.07325440049171447
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,64,power_law_1.01,0.07671679854393006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,64,power_law_1.01,0.08118399977684021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,64,power_law_1.01,0.08024320006370544
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,64,power_law_1.01,0.08477439880371093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,64,power_law_1.01,0.08448640108108521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,64,power_law_1.01,0.08720639944076539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,64,power_law_1.01,0.06673920154571533
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,64,power_law_1.01,0.08680959939956664
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,64,power_law_1.01,0.06325759887695312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,64,power_law_1.01,0.08421760201454162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,64,power_law_1.01,0.061766397953033444
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,64,power_law_1.01,0.08938239812850952
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,64,power_law_1.01,0.05901439785957337
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,64,power_law_1.01,0.08941439986228943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,64,power_law_1.01,0.061350399255752565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,64,power_law_1.01,0.0973695993423462
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,64,power_law_1.01,0.061375999450683595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,64,power_law_1.01,0.09612159729003907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,64,power_law_1.01,0.06527360081672669
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,64,power_law_1.01,0.09815679788589478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,64,power_law_1.01,0.06694399714469909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,64,power_law_1.01,0.10785280466079712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,64,power_law_1.01,0.06709120273590088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,64,power_law_1.01,0.1149440050125122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,64,power_law_1.01,0.06840959787368775
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,64,balanced,2.130773385365804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,64,power_law_1.01,0.1346560001373291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,64,power_law_1.01,0.06993920207023621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,64,power_law_1.01,0.149619197845459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,64,power_law_1.01,0.06949120163917541
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,64,power_law_1.01,0.1859071969985962
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,64,power_law_1.01,0.07272319793701172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,64,power_law_1.01,0.07371519804000855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,64,power_law_1.01,0.22119040489196778
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,64,power_law_1.01,0.0801472008228302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,64,power_law_1.01,0.29381120204925537
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,64,power_law_1.01,0.08082559704780579
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,64,power_law_1.01,0.3529088020324707
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,64,balanced,2.7631359100341797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,64,power_law_1.01,0.08730239868164062
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,64,power_law_1.01,0.5036223888397217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,64,power_law_1.01,0.09973120093345642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,64,power_law_1.01,0.6255104064941406
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,64,power_law_1.01,0.11349120140075683
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,64,power_law_1.01,0.7944704055786133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,64,power_law_1.01,0.14099839925765992
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,64,power_law_1.01,1.2525823593139649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,64,power_law_1.01,0.1651520013809204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,64,power_law_1.01,2.384684753417969
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,64,power_law_1.01,0.21191039085388183
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,64,power_law_1.01,0.25541760921478274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,64,power_law_1.01,0.35802879333496096
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,64,power_law_1.01,0.4507584095001221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,64,power_law_1.01,0.6440063953399658
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,64,power_law_1.01,0.8274880409240722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,64,power_law_1.01,1.0215744018554687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,64,power_law_1.01,1.6141183853149415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,64,power_law_1.01,3.202483367919922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,64,power_law_1.01,0.0705407977104187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,64,power_law_1.01,0.06888960003852844
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,64,power_law_1.01,0.0698751986026764
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,64,power_law_1.2,0.08615040183067321
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,64,power_law_1.01,0.058815997838973996
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,64,power_law_1.2,0.08414720296859741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,64,power_law_1.01,0.06321280002593994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,64,power_law_1.2,0.07934079766273498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,64,power_law_1.01,0.06783999800682068
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,64,power_law_1.2,0.07329919934272766
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,64,power_law_1.01,0.06663680076599121
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,64,power_law_1.2,0.0762112021446228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,64,power_law_1.01,0.07040640115737914
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,64,power_law_1.2,0.07802879810333252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,64,power_law_1.01,0.06787840127944947
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,64,power_law_1.2,0.08005759716033936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,64,power_law_1.01,0.07189760208129883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,64,power_law_1.2,0.08478720188140869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,64,power_law_1.01,0.07152640223503112
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,64,power_law_1.2,0.08380799889564514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,64,power_law_1.01,0.074099200963974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,64,power_law_1.2,0.06437119841575623
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,64,power_law_1.2,0.08384640216827392
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,64,power_law_1.01,0.07438079714775085
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,64,power_law_1.2,0.06349440217018128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,64,power_law_1.2,0.0808896005153656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,64,power_law_1.01,0.07457280158996582
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,64,power_law_1.2,0.06037120223045349
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,64,power_law_1.2,0.08675839900970458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,64,power_law_1.01,0.07960320115089417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,64,power_law_1.2,0.05971199870109558
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,64,power_law_1.2,0.08775039911270141
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,64,power_law_1.01,0.08054400086402894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,64,power_law_1.2,0.0607807993888855
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,64,power_law_1.2,0.08672639727592468
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,64,power_law_1.01,0.08543360233306885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,64,power_law_1.2,0.0643392026424408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,64,power_law_1.2,0.09592959880828858
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,64,power_law_1.01,0.09256960153579712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,64,power_law_1.2,0.06567680239677429
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,64,power_law_1.2,0.09834240078926086
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,64,power_law_1.01,0.10026880502700805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,64,power_law_1.2,0.06563199758529663
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,64,power_law_1.2,0.07008640170097351
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,64,power_law_1.2,0.10304000377655029
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,64,power_law_1.01,0.12599680423736573
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,64,power_law_1.2,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,64,power_law_1.2,0.07073280215263367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,64,power_law_1.2,0.11080960035324097
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,64,power_law_1.01,0.14332159757614135
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,64,power_law_1.2,0.0699455976486206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,64,power_law_1.2,0.06616320013999939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,64,power_law_1.2,0.11811200380325318
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,64,power_law_1.01,0.18542720079421998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,64,power_law_1.2,0.06917759776115417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,64,power_law_1.2,0.060140800476074216
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,64,power_law_1.2,0.13793920278549193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,64,power_law_1.01,0.21599359512329103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,64,power_law_1.2,0.07047039866447449
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,64,power_law_1.2,0.05916159749031067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,64,power_law_1.2,0.15809279680252075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,64,power_law_1.2,0.07112320065498352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,64,power_law_1.01,0.3023231983184814
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,64,power_law_1.2,0.06428800225257873
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,64,power_law_1.2,0.20554239749908448
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,64,power_law_1.2,0.07252479791641235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,64,power_law_1.01,0.3641088008880615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,64,power_law_1.2,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,64,power_law_1.2,0.23220479488372803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,64,power_law_1.2,0.07875199913978577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,64,power_law_1.01,0.530463981628418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,64,power_law_1.2,0.06997759938240052
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,64,power_law_1.2,0.3120255947113037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,64,power_law_1.2,0.0821120023727417
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,64,power_law_1.01,0.6589695930480957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,64,power_law_1.2,0.0700160026550293
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,64,power_law_1.2,0.40687999725341795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,64,power_law_1.2,0.08882560133934021
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,64,power_law_1.01,0.8380415916442872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,64,power_law_1.2,0.10058239698410035
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,64,power_law_1.2,0.07103999853134155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,64,power_law_1.2,0.584006404876709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,64,power_law_1.01,1.2718655586242675
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,64,power_law_1.2,0.11664639711380005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,64,power_law_1.2,0.07238399982452393
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,64,power_law_1.2,0.782528018951416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,64,power_law_1.2,0.14470399618148805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,64,power_law_1.2,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,64,power_law_1.2,0.9051327705383301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,64,power_law_1.01,2.6561344146728514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,64,power_law_1.2,0.16741119623184203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,64,power_law_1.2,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,64,power_law_1.2,1.4241472244262696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,64,power_law_1.2,0.0726207971572876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,64,power_law_1.2,0.22007040977478026
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,64,power_law_1.2,0.07881600260734559
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,64,power_law_1.2,0.2665152072906494
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,64,power_law_1.2,2.763225555419922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,64,power_law_1.2,0.08138880133628845
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,64,power_law_1.2,0.3782272100448608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,64,power_law_1.2,0.08662400245666504
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,64,power_law_1.2,0.4889472007751465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,64,power_law_1.2,0.09514880180358887
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,64,power_law_1.2,0.6696832180023193
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,64,power_law_1.2,0.10438400506973267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,64,power_law_1.2,0.9162367820739746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,64,power_law_1.2,0.13215359449386596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,64,power_law_1.2,1.1287232398986817
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,64,power_law_1.2,0.14888960123062134
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,64,power_law_1.2,1.93255672454834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,64,power_law_1.2,0.19167360067367553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,64,power_law_1.2,3.5642879486083983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,64,power_law_1.2,0.2313152074813843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,64,power_law_1.2,0.3121920108795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,64,power_law_1.2,0.4028736114501953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,64,power_law_1.2,0.5700032234191894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,64,power_law_1.2,0.7715775966644287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,64,power_law_1.2,0.9590784072875976
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,64,power_law_1.2,1.467046356201172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,64,power_law_1.2,2.9597375869750975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,32,balanced,0.058575997749964394
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,32,balanced,0.05788266658782959
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,32,balanced,0.05921066800753275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,32,balanced,0.0647573322057724
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,32,balanced,0.06700266897678375
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,32,balanced,0.07891733447710673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,32,balanced,0.09294933080673218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,32,balanced,0.09071999788284302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,32,balanced,0.050399998823801674
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,32,balanced,0.0888159970442454
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,32,balanced,0.052895997961362205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,32,balanced,0.09108266234397888
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,32,balanced,0.051541333397229515
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,32,balanced,0.09361599882443745
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,32,balanced,0.05564799904823303
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,32,balanced,0.09165866176287334
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,32,balanced,0.0524586687485377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,32,balanced,0.060789331793785095
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,32,balanced,0.09576533238093059
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,32,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,32,balanced,0.06569600105285645
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,32,balanced,0.05213333169619242
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,32,balanced,0.0958079993724823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,32,balanced,0.07212266822655995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,32,balanced,0.05949333310127258
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,32,balanced,0.0735093355178833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,32,balanced,0.09941866993904114
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,32,balanced,0.06131733457247416
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,32,balanced,0.07206933200359344
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,32,balanced,0.10207999746004741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,32,balanced,0.0692799985408783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,32,balanced,0.07211733361085255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,32,balanced,0.07541333138942719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,32,balanced,0.10607999563217163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,32,balanced,0.0754559983809789
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,32,balanced,0.07551999886830647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,32,balanced,0.07458666463692983
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,32,balanced,0.11176533500353496
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,32,balanced,0.0761599987745285
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,32,balanced,0.0753119985262553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,32,balanced,0.07528000076611836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,32,balanced,0.11969600121180217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,32,balanced,0.07774400214354198
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,32,balanced,0.079434668024381
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,32,balanced,0.07784533500671387
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,32,balanced,0.08331733445326488
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,32,balanced,0.13408533732096353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,32,balanced,0.08079466720422109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,32,balanced,0.08796800176302592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,32,balanced,0.07950399816036224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,32,balanced,0.09819733103116353
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,32,balanced,0.14804800351460776
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,32,balanced,0.08754666646321614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,32,balanced,0.10359467069307964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,32,balanced,0.08869866530100505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,32,balanced,0.09166399637858073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,32,balanced,0.11735999584197998
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,32,balanced,0.17779199282328287
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,32,balanced,0.09805333614349365
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,32,balanced,0.14217066764831543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,32,power_law_1.01,0.08087679743766785
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,32,balanced,0.10846933722496033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,32,power_law_1.01,0.0908352017402649
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,32,balanced,0.1643893321355184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,32,balanced,0.20339733362197876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,32,power_law_1.01,0.08555520176887513
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,32,balanced,0.12504532933235168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,32,power_law_1.01,0.07360640168190002
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,32,balanced,0.14056000113487244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,32,power_law_1.01,0.07542399764060974
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,32,balanced,0.2071626583735148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,32,power_law_1.01,0.08026880025863647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,32,balanced,0.25966399908065796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,32,balanced,0.17891200383504233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,32,power_law_1.01,0.08254079818725586
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,32,power_law_1.01,0.08590720295906067
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,32,balanced,0.2490666707356771
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,32,power_law_1.01,0.0851967990398407
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,32,balanced,0.20835200945536295
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,32,power_law_1.01,0.08618879914283753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,32,power_law_1.01,0.08885759711265565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,32,balanced,0.31961599985758465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,32,power_law_1.01,0.08911359906196595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,32,balanced,0.3339039882024129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,32,balanced,0.2725013295809428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,32,power_law_1.01,0.08896639943122864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,32,power_law_1.01,0.09195520281791687
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,32,power_law_1.01,0.09742079973220825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,32,power_law_1.01,0.09752320051193238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,32,balanced,0.42288001378377277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,32,balanced,0.3506346543629964
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,32,power_law_1.01,0.10147839784622192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,32,balanced,0.4286613464355469
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,32,power_law_1.01,0.11016960144042968
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,32,power_law_1.01,0.11775360107421876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,32,power_law_1.01,0.13890559673309327
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,32,power_law_1.01,0.1546175956726074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,32,power_law_1.01,0.1914944052696228
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,32,balanced,0.4779680172602336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,32,power_law_1.01,0.22273919582366944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,32,balanced,0.592741330464681
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,32,power_law_1.01,0.2878079891204834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,32,power_law_1.01,0.3682431936264038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,32,balanced,0.5456693172454834
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,32,power_law_1.01,0.5151552200317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,32,power_law_1.01,0.6676352024078369
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,32,power_law_1.01,0.8244671821594238
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,32,power_law_1.01,1.2804096221923829
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,32,balanced,0.6135093371073405
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,32,balanced,0.7657653490702311
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,32,power_law_1.01,2.5307519912719725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,32,balanced,0.6774933338165283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,32,balanced,0.7518400351206461
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,32,balanced,0.9466559886932373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,32,balanced,1.0191413561503093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,32,balanced,1.1565333207448323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,32,balanced,1.4628160794576008
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,32,balanced,1.9888159434000652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,32,balanced,2.267973264058431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,32,balanced,2.877024014790853
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,32,power_law_1.01,0.06761599779129028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,32,power_law_1.01,0.07451519966125489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,32,power_law_1.01,0.06696959733963012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,32,power_law_1.01,0.05780479907989502
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,32,power_law_1.01,0.05974400043487549
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,32,power_law_1.01,0.06515840291976929
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,32,power_law_1.01,0.06617599725723267
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,32,power_law_1.01,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,32,power_law_1.01,0.0699455976486206
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,32,power_law_1.01,0.07212160229682922
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,32,power_law_1.01,0.07179520130157471
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,32,power_law_1.01,0.07529600262641907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,32,power_law_1.01,0.07549440264701843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,32,power_law_1.01,0.07553920149803162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,32,power_law_1.01,0.08373759984970093
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,32,power_law_1.01,0.08705919981002808
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,32,power_law_1.01,0.09229440093040467
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,32,power_law_1.01,0.10613759756088256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,32,power_law_1.01,0.11927039623260498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,32,power_law_1.01,0.14680960178375244
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,32,power_law_1.01,0.16911360025405883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,32,power_law_1.01,0.22048640251159668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,32,power_law_1.01,0.2686527967453003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,32,power_law_1.01,0.3668607950210571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,32,power_law_1.01,0.46798081398010255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,32,power_law_1.01,0.6613183975219726
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,32,power_law_1.01,0.8753664016723632
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,32,power_law_1.01,1.0503999710083007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,32,power_law_1.01,1.7272192001342774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,32,power_law_1.01,3.2745792388916017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,32,power_law_1.2,0.08053119778633118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,32,power_law_1.2,0.08975359797477722
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,32,power_law_1.01,0.07377920150756836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,32,power_law_1.2,0.07836800217628478
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,32,power_law_1.01,0.0763584017753601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,32,power_law_1.2,0.07289599776268005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,32,power_law_1.01,0.0709824025630951
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,32,power_law_1.2,0.07596160173416137
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,32,power_law_1.01,0.05947520136833191
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,32,power_law_1.2,0.08035839796066284
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,32,power_law_1.01,0.06392959952354431
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,32,power_law_1.2,0.08178560137748718
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,32,power_law_1.01,0.06792960166931153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,32,power_law_1.2,0.0821951985359192
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,32,power_law_1.2,0.06876800060272217
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,32,power_law_1.01,0.06647679805755616
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,32,power_law_1.2,0.08458240032196045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,32,power_law_1.2,0.07289599776268005
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,32,power_law_1.01,0.0753920018672943
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,32,power_law_1.2,0.0877888023853302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,32,power_law_1.2,0.06357759833335877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,32,power_law_1.01,0.07300480008125305
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,32,power_law_1.2,0.08537600040435792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,32,power_law_1.2,0.05847039818763733
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,32,power_law_1.01,0.0738431990146637
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,32,power_law_1.2,0.08935040235519409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,32,power_law_1.2,0.061715197563171384
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,32,power_law_1.01,0.07271680235862732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,32,power_law_1.2,0.06208639740943909
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,32,power_law_1.2,0.08863999843597412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,32,power_law_1.01,0.07625600099563598
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,32,power_law_1.2,0.06453120112419128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,32,power_law_1.2,0.0923904001712799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,32,power_law_1.2,0.06995840072631836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,32,power_law_1.2,0.09721599817276001
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,32,power_law_1.01,0.07560960054397584
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,32,power_law_1.2,0.06893439888954163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,32,power_law_1.01,0.07836160063743591
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,32,power_law_1.2,0.09813759922981262
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,32,power_law_1.2,0.07061120271682739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,32,power_law_1.2,0.10403200387954711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,32,power_law_1.2,0.07308160066604615
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,32,power_law_1.2,0.11197439432144166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,32,power_law_1.2,0.07324159741401673
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,32,power_law_1.2,0.12524800300598143
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,32,power_law_1.2,0.07574399709701538
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,32,power_law_1.2,0.1427072048187256
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,32,power_law_1.2,0.08045439720153809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,32,power_law_1.2,0.1579519987106323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,32,power_law_1.2,0.08414720296859741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,32,power_law_1.2,0.1960255980491638
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,32,power_law_1.2,0.08622080087661743
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,32,power_law_1.2,0.24168319702148439
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,32,power_law_1.2,0.09432960152626038
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,32,power_law_1.2,0.3214207887649536
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,32,power_law_1.2,0.10700160264968872
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,32,power_law_1.2,0.39904639720916746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,32,power_law_1.2,0.12216960191726685
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,32,power_law_1.2,0.5812352180480957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,32,power_law_1.2,0.15226240158081056
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,32,power_law_1.2,0.7760128021240235
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,32,power_law_1.2,0.17630079984664918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,32,power_law_1.2,0.8905216217041015
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,32,power_law_1.2,0.225433611869812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,32,power_law_1.2,1.422719955444336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,32,power_law_1.2,0.2796992063522339
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,32,power_law_1.2,0.38807039260864257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,32,power_law_1.2,2.824224090576172
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,32,power_law_1.2,0.4897280216217041
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,32,power_law_1.2,0.6939008235931396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,32,power_law_1.2,0.9181440353393555
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,32,power_law_1.2,1.1935872077941894
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,32,power_law_1.2,1.8767168045043945
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,32,power_law_1.2,3.7246784210205077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,32,power_law_1.2,0.07317119836807251
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,32,power_law_1.2,0.08006399869918823
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,32,power_law_1.2,0.06892799735069274
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,32,power_law_1.2,0.061267197132110596
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,32,power_law_1.2,0.06360960006713867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,32,power_law_1.2,0.06783360242843628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,32,power_law_1.2,0.06951040029525757
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,32,power_law_1.2,0.07274879813194275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,32,power_law_1.2,0.07233920097351074
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,32,power_law_1.2,0.07337599992752075
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,32,power_law_1.2,0.07422720193862915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,32,power_law_1.2,0.07615360021591186
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,32,power_law_1.2,0.07776640057563781
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,32,power_law_1.2,0.07905279994010925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,32,power_law_1.2,0.08353279829025269
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,32,power_law_1.2,0.08638079762458802
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,32,power_law_1.2,0.09105920195579528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,32,power_law_1.2,0.0998528003692627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,32,power_law_1.2,0.11233279705047608
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,32,power_law_1.2,0.13601280450820924
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,32,power_law_1.2,0.15867520570755006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,32,power_law_1.2,0.20149118900299073
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,32,power_law_1.2,0.24392321109771728
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,32,power_law_1.2,0.3287807941436768
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,32,power_law_1.2,0.4284992218017578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,32,power_law_1.2,0.5857088088989257
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,32,power_law_1.2,0.7857728004455566
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,32,power_law_1.2,1.0173439979553223
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,32,power_law_1.2,1.5968128204345704
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,32,power_law_1.2,3.071750450134277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,16,balanced,0.056559999783833824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,16,balanced,0.056976000467936196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,32,power_law_1.01,0.08364800214767457
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,16,balanced,0.06002133091290792
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,16,balanced,0.06513600051403046
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,32,power_law_1.01,0.08506240248680115
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,16,balanced,0.06902933120727539
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,32,power_law_1.01,0.09033600091934205
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,16,balanced,0.08239466448624928
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,16,balanced,0.09790933132171631
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,32,power_law_1.01,0.09930239915847779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,16,balanced,0.10126933455467224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,32,power_law_1.01,0.10882560014724732
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,16,balanced,0.10011200110117595
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,16,balanced,0.09826667110125224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,32,power_law_1.01,0.12955520153045655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,16,balanced,0.09916266798973083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,16,balanced,0.05045866469542185
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,32,power_law_1.01,0.15234559774398804
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,16,balanced,0.05272000034650167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,16,balanced,0.09985599915186565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,16,balanced,0.052757332722345986
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,16,balanced,0.09994666775067647
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,32,power_law_1.01,0.19512319564819336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,16,balanced,0.05978666742642721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,16,balanced,0.10016000270843506
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,16,balanced,0.0637600024541219
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,32,power_law_1.01,0.23383679389953613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,16,balanced,0.10846933722496033
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,16,balanced,0.07030400137106578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,32,power_law_1.01,0.31407999992370605
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,16,balanced,0.07714666426181793
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,16,balanced,0.10748799641927083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,16,balanced,0.07853333155314128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,32,power_law_1.01,0.39297919273376464
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,16,balanced,0.11337066690127055
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,16,balanced,0.07764799892902374
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,16,balanced,0.07612800101439159
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,16,balanced,0.1204159955183665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,32,power_law_1.01,0.554041576385498
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,16,balanced,0.08021333316961925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,16,balanced,0.1272640029589335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,32,power_law_1.01,0.7145279884338379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,16,balanced,0.08141333361466725
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,16,balanced,0.08154666423797607
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,32,power_law_1.01,0.8773056030273437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,16,balanced,0.14629333217938742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,16,balanced,0.08774933218955994
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,32,power_law_1.01,1.4194879531860352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,16,balanced,0.09062932928403218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,16,balanced,0.09193600217501323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,16,balanced,0.16049066185951233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,32,power_law_1.01,2.7356544494628907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,16,balanced,0.1009226640065511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,16,balanced,0.19153066476186117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,16,balanced,0.11392000317573547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,16,balanced,0.12622400124867758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,16,balanced,0.22039467096328735
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,16,balanced,0.15448533495267233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,16,balanced,0.17803200085957846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,16,balanced,0.27898667256037396
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,16,balanced,0.22175999482472739
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,16,balanced,0.35488001505533856
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,16,balanced,0.2655786673227946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,16,balanced,0.3545653422673543
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,16,balanced,0.47607465585072833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,16,balanced,0.4550400177637736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,16,balanced,0.6106666723887125
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,16,balanced,0.6386400063832601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,16,balanced,0.831653356552124
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,16,balanced,0.7598613103230795
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,16,balanced,0.05205333232879639
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,16,balanced,0.052485331892967224
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,16,balanced,0.05731200178464254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,16,balanced,1.0299253463745117
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,16,balanced,0.06148266792297363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,16,balanced,0.06888533135255177
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,16,balanced,0.07877333462238312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,16,balanced,0.09408533573150635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,16,balanced,0.09578133622805278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,16,balanced,0.09397866328557332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,16,balanced,0.09711466232935588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,16,balanced,0.09985066453615825
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,16,balanced,1.1412586371103923
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,16,balanced,0.0969599982102712
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,16,balanced,0.09550933043162028
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,16,balanced,0.0978613297144572
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,16,balanced,0.1060693363348643
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,16,balanced,0.10432533423105876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,16,balanced,0.11358400185902913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,16,balanced,0.11546132961908977
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,16,balanced,0.12380266189575195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,16,balanced,0.14595199624697366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,16,balanced,0.16517866651217142
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,16,balanced,1.5839786529541016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,16,balanced,0.19967466592788696
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,16,balanced,0.23651733001073202
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,16,balanced,0.30687999725341797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,16,balanced,0.3956640164057414
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,16,power_law_1.01,0.07454079985618592
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,16,balanced,2.225482622782389
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,16,power_law_1.01,0.0869376003742218
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,16,balanced,0.540447990099589
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,16,power_law_1.01,0.09704319834709167
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,16,power_law_1.01,0.07264639735221863
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,16,power_law_1.01,0.07685120105743408
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,16,power_law_1.01,0.08268799781799316
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,16,power_law_1.01,0.08579840064048767
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,16,power_law_1.01,0.0899839997291565
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,16,power_law_1.01,0.089055997133255
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,16,balanced,0.6944373448689779
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,16,power_law_1.01,0.09082880020141601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,16,power_law_1.01,0.09239680171012879
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,16,power_law_1.01,0.09445760250091553
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,16,power_law_1.01,0.09296000003814697
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,16,balanced,3.1162185668945312
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,16,power_law_1.01,0.09804800152778625
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,16,power_law_1.01,0.10274560451507568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,16,power_law_1.01,0.10600320100784302
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,16,power_law_1.01,0.10846079587936401
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,16,balanced,0.8568480014801025
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,16,power_law_1.01,0.11989120244979859
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,16,power_law_1.01,0.12747520208358765
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,16,power_law_1.01,0.15058560371398927
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,16,power_law_1.01,0.16987520456314087
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,16,power_law_1.01,0.20206079483032227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,16,power_law_1.01,0.24303998947143554
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,16,power_law_1.01,0.3182336091995239
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,16,power_law_1.01,0.39625599384307864
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,16,power_law_1.01,0.5688320159912109
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,16,power_law_1.01,0.7420608043670655
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,16,power_law_1.01,0.8754048347473145
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,16,power_law_1.01,1.3193984031677246
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,16,balanced,1.3178986708323162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,16,power_law_1.01,2.684832000732422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,16,balanced,2.577722708384196
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,16,power_law_1.01,0.06350079774856568
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,16,power_law_1.01,0.07406719923019409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,16,power_law_1.01,0.074726402759552
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,16,power_law_1.01,0.05815039873123169
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,16,power_law_1.01,0.06250879764556885
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,16,power_law_1.01,0.06727039813995361
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,16,power_law_1.01,0.07009279727935791
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,16,power_law_1.01,0.07239680290222168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,16,power_law_1.01,0.07457919716835022
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,16,power_law_1.01,0.07617279887199402
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,16,power_law_1.01,0.07612800002098083
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,16,power_law_1.01,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,16,power_law_1.01,0.08172159790992736
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,16,power_law_1.01,0.08062719702720642
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,16,power_law_1.01,0.08319360017776489
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,16,power_law_1.01,0.0929535984992981
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,16,power_law_1.01,0.08190720081329346
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,16,power_law_1.01,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,16,power_law_1.01,0.08845440149307252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,16,power_law_1.01,0.06807680130004883
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,16,power_law_1.01,0.0919871985912323
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,16,power_law_1.01,0.07355520129203796
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,16,power_law_1.01,0.10059520006179809
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,16,power_law_1.01,0.07875199913978577
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,16,power_law_1.01,0.1132159948348999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,16,power_law_1.01,0.08298879861831665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,16,power_law_1.01,0.13102079629898072
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,16,power_law_1.01,0.08567039966583252
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,16,power_law_1.01,0.1577023983001709
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,16,power_law_1.01,0.08803840279579163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,16,power_law_1.01,0.1875391960144043
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,16,power_law_1.01,0.08360959887504578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,16,power_law_1.01,0.23535358905792236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,16,power_law_1.01,0.0897599995136261
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,16,power_law_1.01,0.08850560188293458
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,16,power_law_1.01,0.2894079923629761
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,16,power_law_1.01,0.0941760003566742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,16,power_law_1.01,0.39051520824432373
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,16,power_law_1.01,0.09943680167198181
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,16,power_law_1.01,0.4830656051635742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,16,power_law_1.01,0.09801599979400635
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,16,power_law_1.01,0.7233856201171875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,16,power_law_1.01,0.10599039793014527
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,16,power_law_1.01,0.9467007637023925
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,16,power_law_1.01,0.11282559633255004
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,16,power_law_1.01,1.1231679916381836
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,16,power_law_1.01,0.12785279750823975
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,16,power_law_1.01,1.809516716003418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,16,power_law_1.01,0.15047039985656738
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,16,power_law_1.01,0.1710592031478882
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,16,power_law_1.01,3.546752166748047
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,16,power_law_1.01,0.21755518913269042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,16,power_law_1.2,0.07518079876899719
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,16,power_law_1.01,0.26652159690856936
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,16,power_law_1.2,0.08823040127754211
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,16,power_law_1.01,0.36083838939666746
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,16,power_law_1.2,0.0784063994884491
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,16,power_law_1.01,0.438259220123291
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,16,power_law_1.2,0.07425919771194459
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,16,power_law_1.2,0.07237759828567505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,16,power_law_1.01,0.6338624000549317
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,16,power_law_1.2,0.08183040022850037
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,16,power_law_1.01,0.8183103561401367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,16,power_law_1.2,0.08780800104141236
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,16,power_law_1.01,1.0180607795715333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,16,power_law_1.2,0.08917760252952575
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,16,power_law_1.01,1.6266048431396485
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,16,power_law_1.2,0.0889792025089264
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,16,power_law_1.2,0.08935040235519409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,16,power_law_1.01,3.2323646545410156
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,16,power_law_1.2,0.09113600254058837
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,16,power_law_1.2,0.09324799776077271
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,16,power_law_1.2,0.09673600196838379
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,16,power_law_1.2,0.09846400022506714
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,16,power_law_1.2,0.10323840379714966
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,16,power_law_1.2,0.10514559745788574
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,16,power_law_1.2,0.1115839958190918
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,16,power_law_1.2,0.12257920503616333
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,16,power_law_1.2,0.12933759689331054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,16,power_law_1.2,0.06622080206871032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,16,power_law_1.2,0.1550336003303528
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,16,power_law_1.2,0.07523840069770812
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,16,power_law_1.2,0.1747712016105652
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,16,power_law_1.2,0.06174719929695129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,16,power_law_1.2,0.21375999450683594
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,16,power_law_1.2,0.058873599767684935
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,16,power_law_1.2,0.06225919723510742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,16,power_law_1.2,0.2504127979278564
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,16,power_law_1.2,0.06471679806709289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,16,power_law_1.2,0.3455104112625122
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,16,power_law_1.2,0.07103360295295716
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,16,power_law_1.2,0.4394559860229492
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,16,power_law_1.2,0.07372159957885742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,16,power_law_1.2,0.6131711959838867
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,16,power_law_1.2,0.07537919878959656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,16,power_law_1.2,0.7532479763031006
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,16,power_law_1.2,0.07597439885139465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,16,power_law_1.2,0.9926464080810546
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,16,power_law_1.2,0.07617920041084289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,16,power_law_1.2,0.08136320114135742
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,16,power_law_1.2,1.5239168167114259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,16,power_law_1.2,0.08229119777679443
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,16,power_law_1.2,2.8047935485839846
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,16,power_law_1.2,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,16,power_law_1.2,0.09283199906349182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,16,power_law_1.2,0.0932096004486084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,16,power_law_1.2,0.09959040284156799
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,16,power_law_1.2,0.11666560173034668
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,16,power_law_1.2,0.12900480031967163
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,16,power_law_1.2,0.16151679754257203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,16,power_law_1.2,0.18968960046768188
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,16,power_law_1.2,0.2389120101928711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,16,power_law_1.2,0.29706239700317383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,16,power_law_1.2,0.3989824056625366
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,16,power_law_1.2,0.5395071983337403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,16,power_law_1.2,0.7741119861602783
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,16,power_law_1.2,0.9659775733947754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,16,power_law_1.2,1.237504005432129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,16,power_law_1.2,2.075276756286621
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,16,power_law_1.2,3.870809555053711
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,16,power_law_1.2,0.06954879760742187
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,16,power_law_1.2,0.08076159954071045
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,16,power_law_1.2,0.0746944010257721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,16,power_law_1.2,0.06334720253944397
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,16,power_law_1.2,0.0667136013507843
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,16,power_law_1.2,0.07189120054244995
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,16,power_law_1.2,0.07717120051383972
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,16,power_law_1.2,0.08265600204467774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,16,power_law_1.2,0.08487039804458618
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,8,balanced,0.05566399792830149
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,16,power_law_1.2,0.08268160223960877
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,8,balanced,0.05904533465703329
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,16,power_law_1.2,0.08628479838371277
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,8,balanced,0.06101333101590475
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,8,balanced,0.06516266862551372
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,16,power_law_1.2,0.09233919978141784
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,8,balanced,0.07417066891988118
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,16,power_law_1.2,0.08970879912376403
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,8,balanced,0.09281599521636963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,8,balanced,0.1106773316860199
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,16,power_law_1.2,0.09103360176086425
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,8,balanced,0.10982933640480042
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,16,power_law_1.2,0.09932159781455993
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,8,balanced,0.10964799920717876
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,8,balanced,0.10794132947921753
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,16,power_law_1.2,0.09847040176391601
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,8,balanced,0.10848533113797505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,16,power_law_1.2,0.10802559852600098
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,8,balanced,0.11210667093594869
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,8,balanced,0.11161067088445027
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,16,power_law_1.2,0.12182400226593018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,8,balanced,0.1118239959081014
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,16,power_law_1.2,0.13043199777603148
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,8,balanced,0.12058132886886597
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,16,power_law_1.2,0.1542080044746399
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,8,balanced,0.11921599507331848
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,16,power_law_1.2,0.18035839796066283
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,8,balanced,0.12603200475374857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,16,power_law_1.2,0.23198719024658204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,8,balanced,0.138373335202535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,16,power_law_1.2,0.28041601181030273
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,8,balanced,0.14550399780273438
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,16,power_law_1.2,0.38010239601135254
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,16,power_law_1.2,0.474944019317627
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,8,balanced,0.16622400283813477
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,16,power_law_1.2,0.6720320224761963
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,8,balanced,0.18270933628082275
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,16,power_law_1.2,0.9030015945434571
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,16,power_law_1.2,1.0509568214416505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,8,balanced,0.2185759941736857
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,16,power_law_1.2,1.80263671875
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,16,power_law_1.2,3.4995838165283204
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,8,balanced,0.25195733706156415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,8,balanced,0.3264213403065999
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,8,balanced,0.05242133140563965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,8,balanced,0.05301333467165629
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,8,balanced,0.057114665706952415
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,8,balanced,0.0599839985370636
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,8,balanced,0.0710399995247523
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,8,balanced,0.08346666892369588
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,8,balanced,0.09202133615811665
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,8,balanced,0.42186133066813153
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,8,balanced,0.09578667084376018
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,8,balanced,0.09641599655151367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,8,balanced,0.09583466251691182
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,8,balanced,0.09947199622790019
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,8,balanced,0.09823466340700786
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,8,balanced,0.09915199875831604
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,8,balanced,0.1016373336315155
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,8,balanced,0.10893866419792175
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,8,balanced,0.11212266484896342
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,8,balanced,0.5729120175043741
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,8,balanced,0.12352533141771953
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,8,balanced,0.13545067111651102
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,8,balanced,0.14827199776967367
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,8,balanced,0.17715734243392944
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,8,balanced,0.20153067509333292
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,8,power_law_1.01,0.07047680020332336
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,8,power_law_1.01,0.07828480005264282
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,8,balanced,0.7414720058441162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,8,power_law_1.01,0.09524480104446412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,8,balanced,0.25190399090449017
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,8,power_law_1.01,0.06855040192604064
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,8,power_law_1.01,0.08012800216674805
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,8,balanced,0.3011253277460734
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,8,power_law_1.01,0.08839679956436157
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,8,power_law_1.01,0.09148160219192505
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,8,power_law_1.01,0.09800959825515747
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,8,power_law_1.01,0.10104320049285889
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,8,balanced,0.405189315478007
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,8,power_law_1.01,0.10119040012359619
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,8,power_law_1.01,0.10402560234069824
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,8,power_law_1.01,0.10360959768295289
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,8,balanced,0.9246026674906412
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,8,power_law_1.01,0.10522240400314331
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,8,power_law_1.01,0.10840959548950195
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,8,balanced,0.5305333137512207
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,8,power_law_1.01,0.11408640146255493
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,8,power_law_1.01,0.11880320310592651
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,8,power_law_1.01,0.12315520048141479
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,8,power_law_1.01,0.1345919966697693
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,8,power_law_1.01,0.1467967987060547
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,8,power_law_1.01,0.17437440156936646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,8,balanced,0.7390613555908203
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,8,power_law_1.01,0.19720319509506226
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,8,power_law_1.01,0.2377216100692749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,8,power_law_1.01,0.28557438850402833
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,8,power_law_1.01,0.3619584083557129
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,8,power_law_1.01,0.4499328136444092
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,8,power_law_1.01,0.6698431968688965
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,8,power_law_1.01,0.8420672416687012
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,8,balanced,1.3932159741719563
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,8,power_law_1.01,1.0105536460876465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,8,balanced,0.9648906389872233
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,8,power_law_1.01,1.6079423904418946
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,8,power_law_1.01,3.318035125732422
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,8,balanced,1.197541316350301
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,8,power_law_1.01,0.06543999910354614
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,8,balanced,1.8524479866027832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,8,power_law_1.01,0.07060480117797852
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,8,power_law_1.01,0.08451840281486511
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,8,power_law_1.01,0.06114559769630432
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,8,balanced,2.7191200256347656
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,8,power_law_1.01,0.06881279945373535
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,8,power_law_1.01,0.0738752007484436
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,8,power_law_1.01,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,8,power_law_1.01,0.0895359992980957
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,8,power_law_1.01,0.09018880128860474
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,8,power_law_1.01,0.09325439929962158
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,8,power_law_1.01,0.09133440256118774
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,8,power_law_1.01,0.09635840058326721
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,8,power_law_1.01,0.09468799829483032
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,8,power_law_1.01,0.09893119931221009
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,8,power_law_1.01,0.10693119764328003
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,8,power_law_1.01,0.11070079803466797
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,8,power_law_1.01,0.11980160474777221
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,8,power_law_1.01,0.13519999980926514
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,8,power_law_1.01,0.14991359710693358
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,8,power_law_1.01,0.18252160549163818
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,8,power_law_1.01,0.21310079097747803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,8,power_law_1.01,0.2682624101638794
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,8,power_law_1.01,0.332857608795166
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,8,power_law_1.01,0.4567552089691162
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,8,power_law_1.01,0.5768127918243409
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,8,power_law_1.01,0.8109567642211915
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,8,power_law_1.01,1.0498368263244628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,8,power_law_1.01,1.3204992294311524
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,8,balanced,3.6358559926350913
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,8,power_law_1.01,2.023916816711426
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,8,power_law_1.01,4.096748733520508
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,8,power_law_1.2,0.06977279782295227
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,8,power_law_1.2,0.07925760149955749
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,8,power_law_1.2,0.07210239768028259
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,8,power_law_1.2,0.06695039868354798
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,8,power_law_1.2,0.07909119725227357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,8,power_law_1.2,0.08781440258026123
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,8,power_law_1.2,0.08776959776878357
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,8,power_law_1.2,0.0966592013835907
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,8,power_law_1.2,0.09681919813156128
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,8,power_law_1.2,0.10073599815368653
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,8,power_law_1.2,0.1021183967590332
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,8,power_law_1.2,0.1013375997543335
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,8,power_law_1.2,0.10470399856567383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,8,power_law_1.2,0.10671360492706299
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,8,power_law_1.2,0.11398400068283081
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,8,power_law_1.2,0.11603840589523315
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,8,power_law_1.2,0.12467199563980103
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,8,power_law_1.2,0.13706239461898803
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,8,power_law_1.2,0.14935679435729982
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,8,power_law_1.2,0.1730239987373352
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,8,power_law_1.2,0.19330559968948363
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,8,power_law_1.2,0.24163200855255126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,8,power_law_1.2,0.2909503936767578
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,8,power_law_1.2,0.385644793510437
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,8,power_law_1.2,0.5271552085876465
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,8,power_law_1.2,0.6988224029541016
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,8,power_law_1.2,0.06540160179138184
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,8,power_law_1.2,0.9239680290222168
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,8,power_law_1.2,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,8,power_law_1.2,1.1257920265197754
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,8,power_law_1.2,0.06095359921455383
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,8,power_law_1.2,0.0627839982509613
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,8,power_law_1.2,1.714137649536133
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,8,power_law_1.2,0.0692736029624939
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,8,power_law_1.2,3.67407341003418
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,8,power_law_1.2,0.07376639842987061
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,8,power_law_1.2,0.077292799949646
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,8,power_law_1.2,0.08546559810638428
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,8,power_law_1.2,0.09219840168952942
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,8,power_law_1.2,0.08997120261192322
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,8,power_law_1.2,0.0932096004486084
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,8,power_law_1.2,0.0955136001110077
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,8,power_law_1.2,0.09531520009040832
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,8,power_law_1.2,0.09856640100479126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,8,power_law_1.2,0.10690560340881347
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,8,power_law_1.2,0.1091264009475708
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,8,power_law_1.2,0.117740797996521
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,8,power_law_1.2,0.13595520257949828
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,8,power_law_1.2,0.1524415969848633
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,8,power_law_1.2,0.18686720132827758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,8,power_law_1.2,0.21641600131988525
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,8,power_law_1.2,0.2798719882965088
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,8,power_law_1.2,0.33697919845581054
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,8,power_law_1.2,0.45619840621948243
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,8,power_law_1.2,0.6094016075134278
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,8,power_law_1.2,0.8717823982238769
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,8,power_law_1.2,1.180780792236328
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,8,power_law_1.2,1.4048128128051758
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,8,power_law_1.2,2.2314239501953126
TRTLLM,1.2.0rc5,NVIDIA GB200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,8,power_law_1.2,4.554252624511719
